View
6
Download
0
Category
Preview:
Citation preview
A CFD study of the losses in sharp and radiused
orifices with and without inlet cross-flow
Degree Project in Fluid Mechanics, Second Cycle
SG203X
Author: Tien Nguyen
901212-3015
tien@kth.se
Supervisors: Mihai Mihaescu, Royal Institute of Technology
Sara Rabal, Siemens Industrial Turbomachinery
Examiner: Anders Dahlkild, Royal Institute of Technology
17-11-2014
2
Abstract
Increasing efficiency demands on modern gas turbines require higher operating temperatures
which increase the thermal load on the materials. The secondary air system transfers heat
away from the components to ensure that the operating conditions are under the components’
critical values. In the secondary air system, orifices are an essential component, thus a
thorough understanding of the flow behaviour, more specifically the mass flow and pressure
loss, is crucial. Siemens Industrial Turbomachinery initiated an experimental study regarding
the discharge coefficient for sharp and chamfered orifices (Binder, 2013). This work seeks out
to enhance the previous study with radiused orifices through the use of Computational Fluid
Dynamics (CFD). Additionally, inlet cross-flow was studied for radiused orifices.
A literature study was conducted to provide a deeper understanding of the flow behaviour and
to provide a basis for comparison. A mesh verification study was performed to ensure that the
meshes used were adequate, this was done only for flow cases without inlet cross-flow. A
validation study was conducted with the data from Binder (2013) and Hüning (2012) for two
representative cases without inlet cross-flow and one case with inlet cross-flow. The CFD
calculations showed in general a good agreement with the previous work. Furthermore, a
small turbulence model study was conducted and validated with the work of Binder (2013).
85 cases with different orifice length to diameter ratios, radius to diameter ratios and pressure
ratios were run and 34 cases with inlet cross-flow were calculated with varying length to
diameter ratio, radius to diameter ratio and cross-flow ratio. A comprehensive study of the
results was done with comparison with previous articles and correlations, and the obtained
result showed good agreement regarding the flow behaviour and discharge coefficient.
Furthermore, the CFD approach offered more insight into the flow behaviour since the field
variables can be visualized more easily as compared with experimental studies. Finally, all the
discharge coefficients were collected into three correlations, one for the cases without inlet
cross-flow and two for the cases with inlet cross-flow. The correlations have excellent
performance in their defined range but should not be used outside the range and a number of
suggestions are presented for the use of the correlations outside the range.
3
Table of Contents
List of Figures ............................................................................................................. 5
List of Tables ............................................................................................................... 8
Nomenclature .............................................................................................................. 9
1 Introduction ....................................................................................................... 11
2 Earlier studies and current goals .................................................................... 12
3 Literature review .............................................................................................. 13 3.1 Characterizing the losses ............................................................................ 13 3.2 Parameters influencing the losses in orifices ............................................. 14
3.2.1 General description of the flow field .......................................................... 14 3.2.2 The geometry of the orifice ........................................................................ 14
3.2.3 Flow related parameters ............................................................................. 16 3.3 Choking ...................................................................................................... 20 3.4 Hysteresis ................................................................................................... 21 3.5 Correlations for the discharge coefficient .................................................. 21
4 Theoretical background ................................................................................... 23 4.1 Governing equations in fluid mechanics .................................................... 23
4.1.1 Conservation of mass ................................................................................. 23
4.1.2 Conservation of momentum ....................................................................... 23 4.1.3 Conservation of energy .............................................................................. 24
4.1.4 Equation of state ......................................................................................... 24 4.2 Turbulence .................................................................................................. 25
4.2.1 The nature of turbulence ............................................................................ 25
4.3 Turbulence modelling ................................................................................ 25
4.3.1 Reynolds averaged equations ..................................................................... 25 4.3.2 The turbulent boundary layer ..................................................................... 27 4.3.3 Boussinesq assumption .............................................................................. 28
4.3.4 Two equation models ................................................................................. 29
4.3.5 Reynolds Stress Models (RSM) and Large Eddy Simulations (LES) ........ 30 4.4 Numerical Schemes .................................................................................... 31
5 Methodology ...................................................................................................... 33 5.1 Parameters of study .................................................................................... 33 5.2 Geometry .................................................................................................... 34
5.3 Computational mesh ................................................................................... 36 5.3.1 Mesh type I, sharp-edged orifices .............................................................. 36 5.3.2 Mesh type II, radiused orifices ................................................................... 37
5.3.3 Mesh type III, cross-flow ........................................................................... 38 5.4 General Parameters .................................................................................... 38 5.5 Boundary conditions .................................................................................. 39
5.5.1 No cross-flow ............................................................................................. 39
5.5.2 Cross-flow .................................................................................................. 41 5.6 Mesh verification ........................................................................................ 42
5.6.1 No cross-flow ............................................................................................. 42 5.6.2 Cross-flow .................................................................................................. 51
5.7 Solver validation ........................................................................................ 51 5.7.1 Turbulence models ..................................................................................... 52
4
5.7.2 Model validation ........................................................................................ 54
6 Results – Study I, no cross-flow ....................................................................... 58 6.1 General overview ....................................................................................... 58 6.2 Detailed analysis ........................................................................................ 60
6.2.1 L/D and p1/p2 .............................................................................................. 60 6.2.2 L/D and r/D ................................................................................................ 61 6.2.3 r/D and p1/p2 ............................................................................................... 63 6.2.4 L/D, r/D and p1/p2 ....................................................................................... 64 6.2.5 Additional findings ..................................................................................... 65
6.3 Comparison with chamfered orifices ......................................................... 67 6.4 Comparison with correlations in literature ................................................. 68
7 Results – Study II, cross-flow .......................................................................... 71 7.1 General overview ....................................................................................... 72
7.2 Detailed analysis ........................................................................................ 73 7.2.1 L/D and U/cax ............................................................................................. 73 7.2.2 L/D and r/D ................................................................................................ 74
7.2.3 r/D and U/cax............................................................................................... 75 7.2.4 L/D, r/D and U/cax ...................................................................................... 77
7.3 Comparison with data from inlet cross-flow and rotating orifices ............. 78 7.4 Comparison with correlations in literature ................................................. 80
8 Correlation ........................................................................................................ 82 8.1 Correlation- Study I .................................................................................... 82 8.2 Correlations-Study II .................................................................................. 84
8.3 Concluding remarks ................................................................................... 86
9 Conclusion ......................................................................................................... 88
Bibliography ............................................................................................................. 89
............................................................................................................... 91 Appendix A
................................................................................................................ 92 Appendix B
............................................................................................................... 95 Appendix C
5
List of Figures
Figure 3.1: Overview of orifice geometry. Adapted from Idris & Pullen (2005). ................... 14 Figure 3.2: Schematic of flow through an orifice with inlet cross-flow. Adapted from Idris &
Pullen (2005). ........................................................................................................................... 17 Figure 3.3: Vector field of a rotating orifice seen from the orifice's frame of reference
(Dittman et al., 2003). .............................................................................................................. 18
Figure 5.1: Schematic diagram of test rig (Binder et al., 2014). .............................................. 34 Figure 5.2: Geometry of computational domain L/D = 4.0, r/D = 0.08. (a) Study I. (b) Study
II. .............................................................................................................................................. 35 Figure 5.3: L/D = 4.0, r/D = 0. (a) Structured hexahedral mesh. (b) Unstructured tetrahedral
mesh, Hemisphere I. ................................................................................................................. 37
Figure 5.4: L/D = 4.0, r/D = 0. (a) Unstructured tetrahedral mesh, Hemisphere II. (b) Final
mesh. ........................................................................................................................................ 37
Figure 5.5: L/D = 4.0, r/D = 0.08. (a) Structured hexahedral mesh. (b) Final mesh. ............... 37 Figure 5.6: L/D = 4.0, r/D = 0, cross-flow. (a) Structured hexahedral mesh adapted for cross-
flow. (b) Final mesh. ................................................................................................................ 38 Figure 5.7: L/D = 4.0, r/D = 0.08, cross-flow. Unstructured tetrahedral mesh, Cross-flow
cube. ......................................................................................................................................... 38 Figure 5.8: Boundary conditions. L/D = 4.0, r/D = 0.08. ......................................................... 40
Figure 5.9: Boundary conditions, cross-flow. .......................................................................... 41 Figure 5.10: Separation region illustrated by contour plots of negative axial velocity. (a) Mesh
1. (b) Mesh 4. ........................................................................................................................... 44
Figure 5.11: Lines across the orifice. ....................................................................................... 44 Figure 5.12: Axial velocity profile across the orifice at x/L = 0.095. ...................................... 45
Figure 5.13: Turbulent kinetic energy profile across the orifice at x/L = 0.095. ..................... 45
Figure 5.14: Axial velocity profile across the orifice at x/L = 0.952. ...................................... 46
Figure 5.15: Turbulent kinetic energy profile across the orifice at x/L = 0.952. ..................... 46 Figure 5.16: Axial velocity profile across the orifice at x/L = 0.095, Mesh 1. ........................ 47 Figure 5.17: Contour plot of Mach number for Mesh 4, pressure ratio 1.5 ............................. 48 Figure 5.18: Separation region illustrated by negative axial velocity contours. L/D = 0.4, r/D =
0.04, Mesh 5. ............................................................................................................................ 49 Figure 5.19: Axial velocity profile across the orifice at x/L = 0.4. .......................................... 50 Figure 5.20: Turbulent kinetic energy profile across the orifice at x/L = 0.4. ......................... 50 Figure 5.21: Axial velocity profiles across the orifice at x/L = 0.4, Mesh 5. .......................... 51 Figure 5.22: Validation case 1. L/D = 0.4, r/D = 0. ................................................................. 55
Figure 5.23: Recirculation region illustrated by the contour plot of the negative axial
velocities. (a) L/D = 0.4, r/D = 0, p1/p2= 1.3. (b) L/D = 0.4, r/D = 0, p1/p2= 1.65. .................. 55 Figure 5.24: Validation case 2. L/D = 1.5, r/D = 0. ................................................................. 56
Figure 5.25: Validation case 3, L/D = 2.0, r/D = 0, p1/p2. ........................................................ 57 Figure 6.1: (a) Discharge coefficient CD as a function of pressure ratio p1/p2, L/D = 0.4. (b)
Discharge coefficient CD as a function of pressure ratio p1/p2, L/D = 2.0. .............................. 59 Figure 6.2: (a) Discharge coefficient CD as a function of pressure ratio p1/p2, L/D = 4.0. (b)
Discharge coefficient CD as a function of length to diameter ratio L/D, r/D = 0.04. ............... 59 Figure 6.3: (a) Discharge coefficient CD versus the pressure ratio p1/p2, r/D = 0.04. (b)
Discharge coefficient CD versus the length to diameter ratio L/D, r/D = 0.04. ....................... 61 Figure 6.4: Recirculation region illustrated by contour plots of the negative axial velocity. (a)
L/D = 0.4, r/D = 0.04, p1/p2 = 1.4. (b) L/D = 2.0, r/D = 0.04, p1/p2 = 1.4. ............................... 61
6
Figure 6.5 (a) Discharge coefficient CD versus the radius to diameter ratio r/D, p1/p2 = 1.1. (b)
Discharge coefficient CD versus the length to diameter ratio L/D, p1/p2 = 1.1. ....................... 62 Figure 6.6: Recirculation region illustrated by contour plots of the negative axial velocities.
(a) L/D = 0.4, r/D = 0.16, p1/p2 = 1.1. (b) L/D = 0.4, r/D = 0.20, p1/p2 = 1.1. ......................... 62
Figure 6.7 (a) Discharge coefficient CD versus the pressure ratio p1/p2, L/D = 2.0. (b)
Discharge coefficient CD versus the radius over diameter ratio r/D, L/D = 2.0. ...................... 63 Figure 6.8: Recirculation region illustrated by a contour plot of negative axial velocities. L/D
= 2.0, r/D = 0.16, p1/p2 = 1.4. ................................................................................................... 64 Figure 6.9: Discharge coefficient CD versus radius to diameter ratio r/D, L/D = 0.4. ............. 64
Figure 6.10: Recirculation region illustrated by contour plots of negative axial velocities. (a)
L/D = 0.4, p1/p2 =1.4, r/D = 0.16. (b) L/D = 0.4, p1/p2 =1.4, r/D = 0.20. ................................. 65 Figure 6.11: Discharge coefficient CD versus pressure ratio p1/p2, r/D = 0.16. ....................... 65 Figure 6.12: Comparison of unsteadiness of the recirculation region, L/D = 0.4, r/D = 0.16,
p1/p2 = 1.05. (a) Instantaneous axial velocity at 0.01 s. (b) Time-averaged axial velocity. ..... 66
Figure 6.13: Comparison between chamfered and radiused orifices, L/D = 0.4. ..................... 67 Figure 6.14: Comparison between chamfered and radiused orifices, L/D = 2.0-2.14. ............ 68
Figure 6.15: Comparison of data with correlations. (a) L/D = 0.4, r/D = 0.04. (b) L/D = 2.0,
r/D = 0.04. ................................................................................................................................ 69 Figure 6.16: Comparison of data with correlations, L/D = 4.0, r/D = 0.04.............................. 70 Figure 7.1: Static discharge coefficient CD,s. (a) L/D = 2.0. (b) L/D = 4.0 .............................. 72
Figure 7.2: Total discharge coefficient CD,t. (a) L/D = 2.0. (b) L/D = 4.0. .............................. 73 Figure 7.3 (a) Static discharge coefficient CD,s versus the cross-flow ratio U/cax, r/D = 0.08.
(b) Static discharge coefficient CD,s versus the length to diameter ratio L/D, r/D = 0.08. ....... 74
Figure 7.4: Recirculation region illustrated by contour plots of negative axial velocities. (a)
L/D = 2.0, r/D = 0.08, U/cax =0.689. (b) L/D = 4.0, r/D = 0.08, U/cax =0.689. ...................... 74
Figure 7.5: (a) Static discharge coefficient CD,s versus the radius to diameter ratio r/D, U/cax =
0.259. Right: Static discharge coefficient CD,s versus the length to diameter ratio L/D, U/cax =
0.259. ........................................................................................................................................ 75 Figure 7.6: (a) Static discharge coefficient CD,s versus the cross flow ratio U/cax, L/D = 4.0.
(b) Static discharge coefficient CD,s versus the radius to diameter ratio r/D, L/D = 4.0. ......... 76 Figure 7.7: Stagnation region, L/D = 4.0, r/D = 0.30, U/cax = 1.12. (a) Velocity contour plot.
(b) Static pressure contour plot. ............................................................................................... 76
Figure 7.8: Pressure ratio pt,orifice/p2 at x/L = 0.75 and x/L = 1.0 for L/D = 4.0 , r/D = 0.30. .. 77 Figure 7.9: Static discharge coefficient CD,s versus the length to diameter ratio L/D, U/cax =
1.12 ........................................................................................................................................... 77 Figure 7.10: Comparison with Rohde, et al. (1969). (a) L/D = 1.06-2.0, r/D = 0.1951-0.4878.
(b) L/D = 4.0, r/D = 0. .............................................................................................................. 78
Figure 7.11: Comparison with Jakoby, et al. (1997). (a) L/D 2.0-2.66, r/D = 0.08-0.1. (b): L/D
= 2.0-2.66, r/D = 0.30. .............................................................................................................. 79
Figure 7.12: Comparison with Dittman et al. (2003), L/D = 1.25-2.0, r/D = 0-0.20. .............. 80
Figure 7.13: Comparison with Idris and Pullen correlation. (a) L/D = 2.0, r/D = 0-0.30. (b)
L/D = 4.0, r/D = 0-0.30. ........................................................................................................... 81 Figure 8.1: Discharge coefficients given by the correlation versus the discharge coefficients
from the current study. ............................................................................................................. 83 Figure 8.2: Comparison of intermediate variable values with CFD data. (a): L/D = 2.0, r/D =
0-0.08. (b): L/D = 2.0-4.0, r/D = 0.30. ..................................................................................... 84
Figure 8.3: Static discharge coefficients CD,s given by the correlation versus the static
discharge coefficients CD,s from the current study. .................................................................. 85 Figure 8.4: Total discharge coefficients CD,t given by the correlation versus the total discharge
coefficients CD,t from the current study. ................................................................................... 86
7
Figure 8.5: Comparison of intermediate variable values with CFD data. L/D = 2.0, r/D = 0.08-
0.30 (a): Static discharge coefficient CD,s. (b): Total discharge coefficient CD,t. ..................... 86
Figure B. 1: Axial velocity profile across the orifice at x/L = 0.476. L/D = 1.5, r/D = 0. ....... 92 Figure B. 2: Turbulent kinetic energy profile across the orifice at x/L = 0.476. L/D = 1.5, r/D
= 0. ............................................................................................................................................ 92 Figure B. 3: Axial velocity profile across the orifice at x/L = 0.6. L/D = 0.4, r/D = 0.04. ...... 93
Figure B. 4: Turbulent kinetic energy profile across the orifice at x/L = 0.6. L/D = 0.4, r/D =
0.04. .......................................................................................................................................... 93 Figure B. 5: Axial velocity profile across the orifice at x/L = 0.8. L/D = 0.4, r/D = 0.04. ...... 94 Figure B. 6: Turbulent kinetic energy profile across the orifice at x/L = 0.8. L/D = 0.4, r/D =
0.04. .......................................................................................................................................... 94
8
List of Tables
Table 5.1: Orifice geometry, Study I. ....................................................................................... 33 Table 5.2: Orifice geometry, Study II. ..................................................................................... 34 Table 5.3: Dimensions of parts in the computational domain. ................................................. 35 Table 5.4: Solver settings. ........................................................................................................ 39 Table 5.5: Boundary conditions, no cross-flow. ...................................................................... 40
Table 5.6: Boundary conditions, cross-flow. ........................................................................... 42 Table 5.7: Mesh verification. L/D = 1.5, r/D = 0, mesh parameters. ....................................... 43 Table 5.8: Average element volume comparison normalized with the element volume of Mesh
3. ............................................................................................................................................... 43 Table 5.9: Mesh verification. L/D = 1.5, r/D = 0, results. ........................................................ 43
Table 5.10: Relative difference between Mesh 1 and Mesh 4 in axial velocity and TKE at x/L
= 0.095. ..................................................................................................................................... 47
Table 5.11: Mesh verification. L/D = 0.4, r/D = 0.04, mesh parameters. ................................ 48 Table 5.12: Mesh verification. L/D = 0.4, r/D = 0.04, results. ................................................. 49 Table 5.13: Relative difference between Mesh 5 and Mesh 7 in axial velocity and TKE at x/L
= 0.4. ......................................................................................................................................... 51
Table 5.14: Geometrical and physical parameters for turbulence model study. ...................... 52 Table 5.15: Overview of studied turbulence models. ............................................................... 52
Table 5.16: Relative difference in mass flow with Binder (2013). .......................................... 53 Table 5.17: Summary of validation cases. ............................................................................... 54 Table 6.1: Summary of flow cases in Study I. ......................................................................... 58
Table 6.2: Pressure drop across the orifice. ............................................................................. 60 Table 6.3: Flow cases with supersonic regions. ....................................................................... 66
Table 6.4: Maximum and mean relative difference of correlations. ........................................ 70
Table 7.1: Summary of flow cases in Study II. ........................................................................ 71
Table 7.2: Max and mean relative difference of correlations .................................................. 81 Table 8.1: Overview of correlations. ........................................................................................ 87
Table A. 1: Overview of the meshes used in chapters 6 and 7................................................. 91
Table C. 1: Coefficients aijk. ..................................................................................................... 95 Table C. 2: Coefficients bijk. ..................................................................................................... 95 Table C. 3: Coefficients cijk ...................................................................................................... 96
9
Nomenclature
A0 Orifice area [m2]
A Area [m2]
aijk,bijk, cijk Correlation constants [-]
B Constant [-]
cax Ideal axial velocity based on static
pressure ratio [m/s]
CD Discharge coefficient [-]
cp Heat capacity at constant pressure [J/K]
cv Heat capacity at constant volume [J/K]
Cμ Model constant [-]
Cε1 Model constant [-]
Cε2 Model constant [-]
CFD Computational Fluid Dynamics
D Orifice diameter [m]
Et Total energy per unit volume [J/m3]
e Internal energy per unit volume [J/m3]
f Darcy friction factor [-]
f Body force per unit mass vector [N/kg]
g Variable
(g) Gauge
H Total enthalpy per unit volume [J/m3]
k Turbulent kinetic energy [m2/s
2]
L Orifice length [m]
M Mach number [-]
Mass flow [kg/s]
p Pressure [Pa]
Δp Pressure loss [Pa]
PLC Pressure loss coefficient [-]
Q External heat addition per unit volume
[J/m3]
q Heat flux vector [W/m2]
r Radius of inlet radiusing [m]
R Specific gas constant [J/kg/K]
Re Reynolds number [-]
RMS Root mean square
T Temperature [K]
Δt Time scale [s]
TKE Turbulent kinetic energy
Tu Turbulence intensity [%]
U Cross-flow velocity [m/s]
u Velocity vector [m/s]
u+
Dimensionless velocity [-]
Friction velocity [m/s]
V Velocity vector magnitude [m/s]
w Chamfer depth [m]
wax Ideal axial velocity based on static
pressure ratio [m/s]
10
x Orifice axial distance [m]
y (Orifice) wall distance [m]
y+
Non dimensional wall distance [-]
z Orifice wall distance [m]
Greek symbols
α Model constant [-]
β, β' Model constant [-]
γ Heat capacity ratio [-]
ΓT Turbulent diffusivity [Pa∙s]
δ Boundary layer thickness [m]
δij Kronecker delta, δij = 1 if i=j, else δij = 0
δv Viscous length-scale [m]
ε Turbulent eddy dissipation [m2/s
3]
Velocity head ratio [-]
Von Kármán constant [-]
λ Thermal conductivity [W/m/K]
μ Dynamic viscosity [Pa∙s]
μT Turbulent viscosity [Pa∙s]
ν Kinematic viscosity [m2/s]
Πij Stress tensor
ρ Density [kg/m3]
ζε, ζk, ζω Model constant [-]
τ Viscous stress tensor
ηw Wall shear stress [Pa]
θ Scalar variable
ω Large scale eddy frequency [s-1
]
Superscripts
Degree Mean
Fluctuating
* Modified
Subscripts
1 Upstream of orifice
2 Downstream of orifice
abs Absolute
avg Average
crit Critical
eff Effective
ideal Isentropic process
i,j,k Index notation
major Major
minor Minor
r Radiused
rel Relative
rot Rotation
s Static
t Total
11
1 Introduction
The demand for higher thermal efficiency in gas turbines results in higher operating
temperatures; consequently the materials are subjected to an increased thermal load which can
exceed the critical values and affect the component lifetime negatively. As a result, heat must
be transferred away through an internal flow network commonly named the secondary air
system. Cooling air is bled from the compressor and fed into the secondary air system. Flow
restrictors, such as circular orifices are used to control the flow to ensure a correct flow
distribution in the secondary air system. The challenge is to minimize the bleed air from the
compressor since it does not contribute to the turbine output while it still requires
compressional work. This requires a thorough understanding of the flow through the passages
and orifices in the secondary air system to correctly distribute the cooling air since an
imbalanced flow distribution might lead to premature component failure.
Orifices are commonly employed in the secondary air system and the objective of this work is
to investigate the flow behaviour, more specifically the mass flow and pressure losses, of
orifices through the use of Computational Fluid Dynamics (CFD). The work is a continuation
of the work by Binder (2013) and (Binder et al., 2014) and was conducted for Siemens
Industrial Turbomachinery, Finspång Sweden.
12
2 Earlier studies and current goals
Binder (2013) conducted an experimental study on a stationary engine-like test rig at Siemens
Industrial Turbomachinery, Finspång Sweden, to investigate the effects of how different
parameters, physical and geometrical, influence the flow through the orifices found in
secondary air systems. It was deduced after gathering all the relevant parameters and using
Buckingham π analysis that three dimensionless groups of interest were feasible for the
experimental study, the length to diameter ratio of the orifice L/D, the chamfering depth to
diameter ratio of the orifice w/D and the pressure ratio p1/p2. The study produced a correlation
for the discharge coefficient based on the dimensionless groups. In real life applications,
radiused orifices are also found in turbomachinery, thus the goal with the current thesis is to
supplement the previous work with data from CFD calculations regarding radiused orifices.
This introduces a new dimensionless parameter radius to diameter ratio r/D. Additionally, the
effect of inlet cross-flow is investigated for sharp-edged and radiused orifices. Consequently a
new dimensionless parameter cross-flow ratio U/cax is introduced.
The experimental work was presented in an article (Binder et al., 2014) and references will be
made to the article whenever possible due to the accessibility of the article compared to the
diploma work (Binder, 2013).
13
3 Literature review
3.1 Characterizing the losses
When designing a pipe for internal flow, it is crucial to know the pressure losses arising in the
flow. The pressure losses can be divided into two categories, major and minor losses. Major
losses are the losses that occur due to the viscosity in the fluid and the process is irreversible.
The expression for the pressure loss is given in Equation ( 3.1 ) without derivation (Çengel &
Cimbala, 2006).
( 3.1 )
The Darcy friction factor can be obtained for example in a Moody chart. This pressure loss
can also be expressed as a head loss which is the extra static head hL needed to overcome the
pressure losses. This is illustrated in Equation ( 3.2 ).
( 3.2 )
Minor losses are the pressure losses that arise due to sudden changes in geometry such as
inlets, bends and exits for example. The flow separates and additional turbulent mixing arises
which induces an irreversible pressure loss. An example is when a flow enters a sudden
contraction; the flow will accelerate and separate at the smallest cross-sectional area, also
known as the vena contracta. According to Bernoulli’s principle, an acceleration of flow
velocity will yield a decrease in static pressure. When the flow decelerates after the vena
contracta, the static pressure increase according to the same principle but some of the pressure
energy has been lost due to turbulent mixing. This pressure loss can be expressed as a
pressure loss coefficient PLC which is defined as:
( 3.3 )
It can be seen as the quotient of the difference in pressure energy per unit volume divided by
the kinetic energy per unit volume.
An alternative way to quantify the pressure losses, both major and minor losses, is to define a
compressible discharge coefficient which is the ratio of the mass flow divided by the ideal
mass flow that is achieved in an isentropic process.
( 3.4 )
Applying isentropic relations to find the ideal mass flow, the discharge coefficient can be
expressed as:
√
√ [
]
( 3.5 )
The advantage in utilizing the discharge coefficient is that only the mass flow, pressure and
temperature need to be measured whereas the pressure loss coefficient PLC requires the
14
measurement of velocity and density which is more complicated since the quantities are not
constant for compressible flows and the point of evaluation of these quantities is not specified
(Binder, 2013). Furthermore the compressible discharge coefficient embodies both major and
minor losses in one expression. For this work, the compressible discharge coefficient CD will
be the quantity used for evaluating the losses. The compressible discharge coefficient will
henceforth be referred to as the discharge coefficient and all exceptions will be explicitly
stated.
3.2 Parameters influencing the losses in orifices
There are several parameters, both geometrical and physical, which affect the pressure losses
and consequently the discharge coefficient. A review of the available literature is given in
regard to the parameters that influence the discharge coefficient.
3.2.1 General description of the flow field
Flow entering orifices will separate at the inlet due to the sharp turn experienced by the flow.
The amount of separation depends both on geometrical parameters but also on physical
parameters, this also applies to whether the flow will reattach to the orifice wall or stay
separated to the exit. If full reattachment occurs, a recirculation zone is created where flow is
entrapped (Ward-Smith, 1979). In case where no reattachment occurs, the recirculation zone
is not enclosed but open to the outlet of the orifice and flow is drawn into the recirculation
region (Lichtarowicz et al., 1965).
3.2.2 The geometry of the orifice
Orifices come in all forms and shapes; they can be round or rectangular, have a straight,
chamfered or radiused inlet/outlet, and have a varying diameter to mention some properties.
In this work, round orifices with constant diameter with and without radiusing are studied. To
be able to generalize the study of orifices, the various parameters can be described by
dimensionless groups. The length to diameter ratio L/D describes the ratio between the axial
length L and the orifice diameter D. The chamfering depth to diameter ratio w/D describes the
ratio between the axial depth of the chamfering and the orifice diameter. The chamfering can
have different angles; Binder et al. (2014), which this study connects to, used a chamfering
angle of but other chamfering angles have been studied (Hay & Spencer, 1992). The
radius to diameter ratio r/D describes the rounding radius to diameter ratio. A schematic of
these geometrical parameters can be seen in Figure 3.1.
Figure 3.1: Overview of orifice geometry. Adapted from Idris & Pullen (2005).
15
3.2.2.1 Length to diameter ratio L/D
Lichtarowicz et al. (1965) concluded that the incompressible discharge coefficient rises
sharply for with a further increase to and then decreases almost
linearly as L/D becomes larger. They observed that full reattachment occurs for L/D = 4.0.
Deckker & Chang (1965) found, through the use of a static pressure traverse, evidence that for
L/D = [1.0,2.0], pressure recovery existed for compressible flow. Hüning (2008) states that for
L/D > 2, flow reattachment occurs which enables pressure recovery and increases CD as
compared with short orifices without flow reattachment.
3.2.2.2 Chamfer depth to diameter ratio w/D
Orifices can be manufactured with a chamfer usually of 45 or 30 , but other angles do exist.
The chamfer helps the flow to enter the orifice axially thereby reducing the separation and
consequently the inlet losses. Hay & Spencer (1992) found experimentally that the discharge
coefficient increased up to 22%, for a pressure ratio of 1.6 and for orifices with 0.25 ≤ L/D ≤
2.0 with a chamfering angle of 45 . At least 95% of the increase (22%) occurred at w/D =
0.04. The size of the vena contracta appears constant for w/D ≥ 0.08 and therefore higher
chamfering is not more beneficial for the discharge coefficient. Also, a chamfering angle of
30 showed to be more beneficial for the discharge coefficient in comparison with the 45 chamfer angle. It exhibited largely the same behaviour as its counterpart but with a
systematically larger discharge coefficient when compared for different values of L/D. Binder
et al. (2014) found a non-linear increase of the discharge coefficient when increasing w/D;
this was more pronounced for L/D < 2.14. Binder et al. (2014), Dittman et al. (2003) and Hay
& Spencer (1992) observed that for L/D ≤ 0.5, excessive chamfering has negative effect on
the discharge coefficient. Hay & Spencer (1992) proposed that this is due to marginal
reattachment. The effective length of the orifice is reduced when the chamfer depth is
increased, thereby influencing the possibility of reattachment and consequently pressure
recovery which is associated with an increase in discharge coefficient. Binder et al. (2014)
therefore proposed that the effective length of the orifice should be considered, defined as:
( 3.6 )
Applying the definition, Binder et al. (2014) recommend from their measurements a Leff >
0.35 to avoid the decrease in discharge coefficient. They apply the same definition on the data
by Dittman et al. (2003) and obtains a similar value of Leff > 0.3. However, when applying
this to the data of Hay & Spencer (1992), the value of the smallest Leff range from 0.05 to 0.8.
3.2.2.3 Radiusing to diameter ratio r/D
Radiusing has a similar effect on the discharge coefficient as chamfering, i.e. the discharge
coefficient is increased when applying radiusing. Hay & Spencer (1992) observed
experimentally that for L/D = 0.25, small chamfering with a chamfer angle of 45° is superior
to small radiusing but noticed that large radiusing is more beneficial than large chamfering.
However, by using an orifice with a 30° chamfer angle, the discharge coefficient was equal
for both chamfering and radiusing for high values of w/D and r/D. They also noted that that
the chamfered orifices have less dependence on the pressure ratio than the radiused orifices.
Dittman et al. (2003) observed an increase of the discharge coefficient up to 21% for r/D =
0.05 and an increase up to 39% for r/D = 0.20 compared with a sharp-edged orifice with r/D =
0.
16
3.2.2.4 Surface roughness
No study has been found where the author(s) study the effect of wall roughness. However,
there are several studies where the effects of long orifices are shown. Lichtarowicz et al.
(1965) present a linearly decreasing incompressible discharge coefficient for 2.0 ≤ L/D ≤
10.0. This is presumably due to frictional losses, i.e. major losses. For compressible flows,
Fanno friction effects start to become prominent for L/D > 2.0 leading to a decreasing
discharge coefficient according to Parker & Kercher (1991). However in a review by Hay &
Lampard (1996), the effect of friction and consequently surface roughness, is stated to be
negligible for L/D < 8.0.
3.2.3 Flow related parameters
3.2.3.1 Pressure ratio p1/p2
The pressure ratio, denoted as p1/p2, is the ratio of orifice upstream static (stagnation) pressure
to the orifice downstream static pressure. Deckker & Chang (1965) found that the discharge
coefficient increases when the pressure ratio increases and the dependence is strong for
pressure ratios up to 2. This was found to be more pronounced for short orifices. The same
behaviour was also observed by Brain & Reid (1975), Hay & Spencer (1992) and Binder et al.
(2014). Ward-Smith (1979) explained this as an effect of the compressibility of flow for high
pressure ratios. When the pressure ratio increases, the size of the recirculation zone is
suppressed (i.e. the area of the vena contracta increases) and the amount of separated flow
decreases which yields a higher discharge coefficient.
3.2.3.2 Reynolds number Re
The Reynolds number, denoted as Re, is an important quantity in fluid mechanics. It expresses
the ratio of inertial forces to viscous forces and is defined as:
( 3.7 )
Lichtarowicz et al. (1965) gives a recommendation that the incompressible discharge
coefficient is constant for Re > 2∙104 for 2 < L/D <10. Deckker & Chang (1965) found
experimentally that the compressible discharge coefficient is unaffected by the Reynolds
number if the latter being larger than 104.
3.2.3.3 Inlet cross flow
Inlet cross flow introduces a velocity perpendicular to the orifice; a schematic can be seen in
Figure 3.2.
17
Figure 3.2: Schematic of flow through an orifice with inlet cross-flow. Adapted from Idris & Pullen (2005).
A new parameter is introduced, namely the ratio of the cross flow velocity perpendicular to
the orifice, to the ideal axial velocity in the orifice wax. This is based on the total pressure
upstream and given as (Hüning, 2012):
√[ (
)
] [ (
)
]
( 3.8 )
Another way of defining the cross flow ratio is by using ideal axial velocity cax which is not
based on the total pressure but rather the total pressure in the orifice direction, which is here
equivalent to the static pressure (McGreehan & Schotsch, 1988). Equation ( 3.8 ) then
becomes:
√[ (
)
] [ (
)
]
( 3.9 )
It might be noted that the cross flow velocity U is the ideal cross-flow velocity instead of the
actual cross-flow velocity.
Two counteracting effects generally arise when cross-flow at the inlet is present (Hüning,
2008):
1. The flow separation region in the orifice on the upstream flow side is augmented
which leads to a lower through flow due to the reduction of the effective flow area.
2. The occurrence of flow stagnation in the orifice inlet which enables a partial recovery
of the dynamic pressure from the cross-flow. The actual pressure that drives the flow
through the orifice has a value between the total pressure in the orifice’s direction and
the total pressure where the cross-flow dynamic pressure is included, this leads to an
augmentation in orifice through flow.
There are two definitions of the discharge coefficient when inlet cross flow is present. The
discharge coefficient can be defined without the inclusion of the tangential velocity
18
component in the total pressure and temperature in equation ( 3.5 ) but rather the total
quantities along the axis of the orifice, in this case equal to the static pressure and temperature
(McGreehan & Schotsch, 1988). The discharge coefficients based on the static values for
cross flows will be denoted as CD,s:
√
√ [
]
( 3.10 )
However, in many publications with cross flows where the orifices are inclined, the total
quantities are used (Rohde et al., 1969; Hüning, 2012). This definition of the discharge
coefficient will be denoted as CD,t:
√
√ [
]
( 3.11 )
Hüning (2012) found the total discharge coefficient CD,t to decrease for sharp-edged orifices
with increasing cross flow ratios. He also noted a slight raise of the discharge coefficient with
increasing static pressure ratio and also an augmentation of the discharge coefficient with
increasing L/D. The results obtained were similar to the results by Rohde et al. (1969) and
Dittman et al. (2003).
3.2.3.4 Rotation
Rotation is not studied in the current work but due to the fact that it is fairly similar to inlet
cross flow, rotation will be treated as well. Focus will be on rotation with an axis parallel to
the orifice axis.
Figure 3.3: Vector field of a rotating orifice seen from the orifice's frame of reference (Dittman et al., 2003).
From the orifice’s point of view, inlet and exit cross-flow is seen due to the rotation, this is
illustrated by Figure 3.3. Similarly to orifices with inlet cross flow, there exist two definitions
of the discharge coefficient. In the absolute frame, the absolute discharge coefficient CD,abs
does not account for the rotational effects from a rotating orifice and only depends on total
19
values upstream and static values downstream of the orifice. Its form is identical to Equation (
3.5 ). To account for the rotational effects, a relative discharge coefficient CD,rel can be used
which is based on the total pressure in the rotating (relative) frame, i.e. the frame of the
rotating orifice. The relative total temperature and relative total pressure are given as:
( 3.12 )
(
)
( 3.13 )
Urot is the tangential velocity of the orifice, consequently the orifice should experience a
cross-flow velocity Urot. The relative discharge coefficient can now be expressed as
(Dittman et al., 2003):
√
√
[
]
( 3.14 )
The cross-flow ratio found in Equation ( 3.8 ) for inlet cross-flow is written in the relative
frame for a rotating orifice as:
√ [ (
)
]
( 3.15 )
Analogue with the previous example, the cross-flow ratio found in Equation ( 3.9 ) is for
rotating orifices written in the absolute frame as (Dittman et al., 2003):
√ * (
)
+
( 3.16 )
Hüning (2008) discusses the additional effects from centrifugal forces that are present for a
rotating inlet flow or a rotating orifice but disregards this effect when using data from rotating
orifices by Dittman et al. (2003) to present the effect of inlet cross-flow. Hüning (2012)
experimentally studied inlet cross flow for a stationary orifice and stated that direct
comparison can be made between the discharge coefficient based on total pressure ratio for
inlet cross flow CD,t in Equation ( 3.11 ) with the relative discharge coefficient CD,rel given in
Equation ( 3.14 ) when comparing his results with the data from Dittman et al. (2003) based
on the discussion in his previous work (Hüning, 2008). This statement implicitly assumes that
the definition of cross-flow velocity ratio as seen in Equations ( 3.8 ) and ( 3.15 ) can be
compared directly as well. Applying the same reasoning as Hüning, the static discharge
coefficient defined in Equation ( 3.10 ) should be to some extent comparable to the
discharge coefficient in the absolute frame as defined in equation ( 3.5 ) since neither
of them consider the effect of cross-flow as seen by the orifice whether the orifice is
stationary or rotating. Also, the cross-flow velocity ratio as defined in equation ( 3.16 ) should
then be comparable to the cross-flow velocity ratio as defined in equation ( 3.9 ).
20
The effect of an axially rotating orifice has been studied by several authors. Jakoby et al.
(1997) found that the absolute discharge coefficient CD,abs decreased for sharp orifices when
increasing the rotational velocity since the flow becomes more separated in the orifice causing
a decrease in the area of the vena contracta. However, an increase in the absolute discharge
coefficient was observed for an orifice with L/D = 2.66, r/D = 0.5, with increasing rotational
velocity but also an increase followed by a decrease for an orifice with L/D = 2.66, r/D = 0.3.
This was explained as the effect of work transfer from the rotating orifice to the flow. If the
orifice is long enough, the flow will exit parallel to the orifice axis. From the point of view of
the rotating orifice, the flow has been deflected and work has then been done on the flow
according to Euler’s pump equation. Thus two counteracting effects are present: The
separation decreases the discharge coefficient whereas the rotation enhances the discharge
coefficient but the latter is dependent on inlet geometry, being pronounced for large inlet
radiusing. Dittman et al. (2003) studied the effect of chamfering and radiusing on the
discharge coefficient for rotating orifices and observed similar trends to Jakoby et al. (1997).
Hüning (2008) explained that the rotational work transfer as stated earlier can simply be seen
as the “impact of flow stagnation in the relative system” thus, as understood by the author,
enabling the comparison between orifices with an inlet cross-flow and rotating orifices.
3.3 Choking
Choked flow is the condition where changes in downstream pressure do not affect the flow
field upstream. For one dimensional isentropic flow, the local axial velocity will reach sonic
conditions at the smallest cross sectional area, being it a physical wall in a nozzle or a “wall”
formed by a separation bubble. The “wall” formed by the separation bubble, commonly called
vena contracta, acts as a solid wall as no information can propagate upstream. Gas dynamics
gives the relation between total pressure and static pressure as:
(
)
( 3.17 )
For air, with the value of , it can be deduced that for a Mach number larger than
unity, the pressure ratio has to be larger than 1.8929, this is denoted as the critical
pressure ratio. The flow is choked over this critical pressure ratio. The critical isentropic mass
flow, is defined as:
√
(
)
( 3.18 )
If the upstream total pressure is held constant, the mass flow is limited for all pressure ratios
above critical pressure ratio if the walls are physical as the smallest cross-sectional area
cannot be altered. For orifices where the separation of flow forms a vena contracta, the
choking behaviour is different due to the fact that the size of the separation bubble can be
altered by varying the downstream pressure. Also depending on the orifice geometry, choking
can occur at the exit of the orifice. An excellent treatment of this subject can be found in
Ward-Smith (1979).
The choked flow regime involves regions with transonic flow which presents an additional
phenomenon, namely shocks. To capture shocks accurately with CFD, the computational
mesh must be modelled with a sufficiently fine mesh at the shock region to accurately capture
21
the shock discontinuity. Due to time constraints, this was not investigated in this thesis;
therefore this part will purposefully be kept brief.
The concept of the one-dimensional flow is an idealized treatment of flows in orifices and
somewhat incomplete. Weir et al. (1956) found, using Schlieren visualization, oblique shock
waves in two-dimensional orifices, and not the normal shock usually associated to the one-
dimensional flow theory.
Deckker & Chang (1965) observed that for a very thin orifice, L/D ~ 0, the flow never chokes.
Brain & Reid (1975) found that the flow does not choke for an orifice with L/D = 0.14 up to a
pressure ratio of 5, furthermore they found choking to occur above the critical pressure ratio
for L/D = [0.28,0.31]. Similarly Binder et al. (2014) found that choking for L/D = 0.5 occurs
above the critical pressure ratio. On the other hand, Brain & Reid (1975) found that for L/D >
1, choking occurred below the critical pressure ratio. This was thought to be due to the
pressure recovery in the orifice. The same behaviour for orifices with L/D between 0.75 and
2.14 was observed by Binder et al. (2014).
3.4 Hysteresis
Several studies have found evidence of hysteresis in the discharge coefficient depending on
whether the pressure ratio was increased or decreased.
Deckker & Chang (1965) found evidence of hysteresis for L/D = 0.5 for sharp-edged orifices
near the choking region; it was clearly manifested as two closed branches. The phenomenon
was attributed to the attachment and detachment of the jet to and from the wall. Brain & Reid
(1975) also observed hysteresis for sharp-edged orifices with L/D = 0.5 but not for orifices
with L/D < 0.5. They also attributed this to jet reattachment. Binder et al. (2014) did not
experience hysteresis for a sharp-edged orifice with L/D = 0.5 but experienced it when having
a slight chamfering (w/D = 0.00943 and w/D = 0.0169). Furthermore, hysteresis was also
observed for a sharp-edged orifice with L/D = 0.75 which has never been noticed before.
Furthermore, they also observed hysteresis for low pressure ratios depending on whether the
flow valve was opening or closing, i.e. whether the pressure ratio was increased or decreased.
However, the findings are not consistent and they call for more investigations before
hysteresis can be confirmed. The findings in this subchapter have all been found
experimentally near the choking region and since this region is not treated in the current
study, the effect of hysteresis will not be an issue. It is nonetheless good to have the
knowledge of the hysteresis phenomenon associated with flow through orifices.
3.5 Correlations for the discharge coefficient
There are several correlations for the discharge coefficient. Lichtarowicz et al. (1965)
provided a correlation for the incompressible discharge coefficient which regards length to
diameter ratio and the Reynolds number. Parker & Kercher (1991) presented a correlation for
the compressible discharge coefficient which included the effects of Reynolds number,
radiusing and length to diameter ratio. Idris & Pullen (2005) provided an enhanced correlation
including the effects of inlet chamfering, rotation and pumping effects due to rotation with an
inclined orifice. Hüning (2008) provided a similar correlation to the one by Parker & Kercher
(1991) where chamfering and inlet cross-flow was added. All previous correlations calculate
the effect of the different parameters in an additive approach, where each parameter has a
separate term. However, Binder et al. (2014) presented a correlation which includes the
22
pressure ratio, length to diameter ratio and inlet chamfer to diameter ratio where the effects of
the parameters on each other are taken into account.
23
4 Theoretical background
In this chapter, the governing equations in fluid mechanics are presented; the first subchapter
will treat the formulations of the governing equations. The second and third subchapter will
present turbulence and subsequently its modelling. The final subchapter will briefly treat the
numerical methods used for solving the governing equations in the current flow solver CFX.
4.1 Governing equations in fluid mechanics
In fluid mechanics, the governing equations are based on conservation laws, namely
conservation of mass, momentum and energy. Additionally, an equation of state is necessary
for joining the thermodynamic quantities pressure, density and temperature. For the interested
reader, the details of the derivation for this subchapter can be found in Tannehill et al. (1997)
4.1.1 Conservation of mass
The conservation of mass law, also known as the continuity equation, can be seen in Equation
( 4.1 ) :
( 4.1 )
The first term describes the density rate of change in a control volume and the second term
describes the mass flux rate through the surface of the control volume.
The material derivative is defined as :
( 4.2 )
Thus the conservation of mass law found in Equation ( 4.1 ) can be written as:
( 4.3 )
For the incompressible flows, the first term vanishes since the density does not vary
(assumption valid for M < 0.3) and one obtains that the divergence of the velocity is equal to
zero.
4.1.2 Conservation of momentum
The conservation of momentum can be derived from Newton’s second law and is given as:
( 4.4 )
The rate of change of momentum in the control volume is described by the first term on the
left hand side. The second term on the left hand side describes the momentum flux through
the surface of the control volume. The body force per unit volume is described by the first
term on the right hand side. The second term on the right hand side describes the surface force
per unit volume applied on the fluid element and consists of shear and normal stresses in the
24
stress tensor Πij. By expanding the second term on the left hand side in Equation ( 4.4 ) and
combining it with the continuity equation, the conservation of momentum can be expressed
as:
( 4.5 )
Fluids that have a linear relation between stress and strain rate are called Newtonian fluids
and the stress tensor Πij can be written, as a function of pressure and velocity, in a tensor form
as:
(
)
( 4.6 )
The second term on the right hand side is usually called the viscous stress tensor τij. The final
form of the momentum conservation equation, also known as the Navier-Stokes equation, can
then be expressed as:
(
)
( 4.7 )
4.1.3 Conservation of energy
The conservation of energy can be found by applying the first law of thermodynamics on an
infinitesimal fixed control volume. This yields the energy equation with Et being the total
energy per unit volume:
( 4.8 )
The terms on the left hand side represent the rate of change of total energy in the control
volume and the total energy flux through the boundaries respectively. The rate of heat from
external sources is represented in the first term on the right hand side. The second term is the
heat flux through the boundaries and can be expressed as:
( 4.9 )
which is known as Fourier's law for heat transfer where λ is the thermal conductivity. The
third and fourth term on the right hand side represents the work done on control volume by
body and surface forces respectively.
4.1.4 Equation of state
The equation system formed by Equations ( 4.3 ), ( 4.7 ) and ( 4.8 ) contain seven scalar
quantities, ρ, p, e, T, u, v, w, when regarding a Cartesian system and disregarding the body
forces and external heat addition. However, the number of equations only amount to five
scalar equations which presents a closure problem. For gases whose intermolecular forces are
negligible, the perfect gas equation of state is valid:
25
( 4.10 )
where R is the specific gas constant. The assumption of a calorically perfect gas can be
assumed for low temperature situations, this yields that the specific heat capacities at constant
volume cv is constant. The internal energy e can be defined as:
( 4.11 )
With these seven equations, the system can now be closed.
4.2 Turbulence
Turbulent flows are all around us in nature, observe for example the smoke from a cigarette or
candle light, it ascends in a smooth manner initially until it suddenly behaves chaotically. In
engineering applications, turbulent flows are prevalent, they occur for example in flows
around vehicles, in pipes, in combustion to mention some areas.
4.2.1 The nature of turbulence
Turbulent flows have a number of characteristics (Kundu & Cohen, 2008):
They are chaotic.
They have enhanced diffusivity of momentum and heat.
They have vortices, also called eddies, with a large scale of sizes.
Most of the energy is found in the largest eddies. The energy is transferred to the
smaller eddies in a process known as the energy cascade. Non-linear viscous diffusion
dissipates the energy in the smallest eddies. Thus a constant supply of energy is
required for turbulent flows to balance the losses from viscous dissipation.
A famous example is the experiments carried out by Osbourne Reynolds. The flow of dye in a
pipe showed that transition from laminar flow to turbulent flow occurs at a Reynolds number
of around 2300-4000 (Pope, 2000).
4.3 Turbulence modelling
There are several ways to model turbulent flows. The current work uses a Reynolds averaged
Navier Stokes (RANS) approach and thus, the focus of the following subchapters will be on
this approach.
4.3.1 Reynolds averaged equations
The continuity, momentum and energy equations are still valid for turbulent flows but since
there are so many scales involved, solving the equations for all the scales presents quite a
challenge for today’s computers. An alternative to solving for all scales is to solve for the
mean flow characteristics averaged in time. This is known as Reynolds averaging which is
obtained by the decomposition of the variables, e.g. g, into a temporal mean component and
26
fluctuating component followed by a time averaging of the governing equations. For
compressible fluids, the density fluctuations have to be taken into account and the time
averaging is weighed by the density, this is the case with the current flow solver CFX 14.5
(Solver Theory Guide, ANSYS Inc, 2012). For simplicity, the simpler Reynolds averaging
will be presented here, as a consequence the small density fluctuations are not taken into
account in the discussion. This subchapter is based on Tannehill et al. (1997) and the CFX
Solver Theory Guide (ANSYS Inc, 2012). The definition of a time averaged mean quantity
is given as:
∫
( 4.12 )
The time scale Δt has to be larger than the fluctuating motions but small compared to the
unsteady flow field time scale. Time-averaging the fluctuating quantity in a similar fashion
as Equation ( 4.12 ) yields . The variables can be written in a decomposed form.
( 4.13 )
Inserting the decompositions into the continuity equation as seen in Equation ( 4.1 ), and
averaging the whole equation, the Reynolds-averaged continuity equation becomes:
( )
( 4.14 )
The momentum equation becomes:
( )
( 4.15 )
The Reynolds-averaged momentum equation looks familiar save the addition of the term
also known as the Reynolds stress tensor. It is a symmetric tensor with nine
components. When regarding the last term in equation ( 4.15 ), the first term describes the
molecular diffusion of momentum while the second term can be taken as the macroscopic
diffusion of momentum caused by turbulent fluctuations with the latter generally being larger
than the former.
The energy equation has been expressed with the total energy Et previously. With the
assumption of internal energy and kinetic energy being the only components of the total
energy, the latter can be replaced with where H is the total enthalpy (
) and h is the enthalpy ( ). Neglecting the rate of heat from external sources
,
Equation ( 4.8 ) can be rewritten as:
( )
( 4.16 )
The total enthalpy H is also decomposed into a mean and fluctuating part and the final
Reynolds averaged resulting energy equation becomes (CFX Solver Theory Guide, ANSYS
Inc, 2012):
27
(
)
( 4.17 )
It may be noted that
where the last term is also known as the
turbulent kinetic energy k. It can be seen that there has been two terms added. In the second
term on the right hand side, the Reynolds stress enters in the viscous work term and in the
middle term in the second equation on the left hand side, the turbulent flux has been
added. These extra terms arising from Reynolds-averaging hands us a slight problem. There
are more unknowns than equations and the equation system cannot be closed.
4.3.2 The turbulent boundary layer
Before tackling the problem with extra terms, it might be interesting to study the turbulent
boundary layer. In wall bounded flows, a viscous fluid cannot slip against the wall, hence
giving rise to a no-slip condition which imposes a zero flow velocity condition at the wall.
This gives rise to a velocity gradient from the wall up to the free stream. One classical
(laminar) example of this is the Blasius solution of the boundary layer on a flat plate where
the velocity profile can be solved numerically (Kundu & Cohen, 2008).
The following part is based on the work of Pope (2000).
In turbulent flows, the boundary layer requires a separate treatment. Near the wall, the stresses
that are given by the viscous stresses and Reynolds stresses, are dominated by the former one.
As a result, the velocity profile is highly dependent on the Reynolds number. It might then be
appropriate to define a viscous scale applicable in the near-wall region. Instead of using the
index notation, Cartesian coordinates will be used for clarity. The friction velocity is
defined, with the help of wall shear stress as:
√
( 4.18 )
A viscous length scale is defined as:
√
( 4.19 )
The distance from the wall can be measured with the viscous length scale, commonly denoted
as which is defined as:
( 4.20 )
For < 50, usually called the viscous wall region, viscosity contributes significantly to the
total shear stress. For > 50, the region is usually denoted as the outer layer and the effect
of viscosity is negligible on the shear stress. For y+ < 5, the region is called the viscous
sublayer and the Reynolds stress has a negligible impact.
Prandtl postulated the existence of an inner layer, given a high Reynolds number, where the
mean velocity is determined not by boundary layer thickness or free stream velocity but by
28
the viscous scales only. The inner layer spans from the wall to
where is the
boundary layer thickness. For y+ < 5, i.e. the viscous sublayer, the dimensionless velocity
can be taken as:
( 4.21 )
This linear relation breaks down when y+ > 12, where the deviation is larger than 25%. For 30
< y+, y/δ < 0.3, a logarithmic velocity profile is valid:
( 4.22 )
The buffer layer is located between the viscous sublayer and the region where the logarithmic
law holds and is the transition between the viscous and turbulent zones.
4.3.3 Boussinesq assumption
This subchapter is based on CFX Solver Theory Guide (ANSYS Inc, 2012).
The Boussinesq assumption relates the Reynolds stress proportionally to the mean rate of
strain:
(
)
(
)
( 4.23 )
Equation ( 4.23 ) also introduces which is the turbulent viscosity. This term will be subject
to modelling and discussed in subsequent subchapters. The Reynolds averaged momentum
equation seen in Equation ( 4.15 ) can finally be written as:
( )
(
)
( 4.24 )
Firstly, the pressure term is modified to include the two last terms in Equation ( 4.23 ):
( 4.25 )
CFX disregards the last term on the right hand side even though it is not fully correct and
equates by default with p. Secondly, the stress term on the right hand side now have an
effective viscosity in which the molecular and turbulent viscosity is included in. For the
energy equation, the turbulent flux is modelled in the analogous way with the
introduction of the turbulent diffusivity :
( 4.26 )
The turbulent diffusivity is given as
by default in CFX. The assumption of a
constant relation between the turbulent viscosity and turbulent diffusivity is valid for a Prandtl
number (momentum diffusivity over thermal diffusivity) larger than 0.5. In this thesis, the
29
fluid is air which has a Prandtl number of approximately 0.7 so the constant relation between
the turbulent viscosity and turbulent diffusivity is a valid assumption (Schlichting & Gersten,
2006). The energy equation found in Equation ( 4.17 ) can, with the Bossinesq assumption,
now be written as:
(
)
( 4.27 )
4.3.4 Two equation models
To solve the Reynolds averaged equations the turbulent viscosity must be determined in
some manner since it cannot be taken as constant in the flow field.
4.3.4.1 The k-ε model
The model is a two equation model which incorporates two additional transport
equations, one for the turbulent kinetic energy and one for the dissipation rate of turbulent
kinetic energy . With these two quantities, the turbulent viscosity (and the turbulent
diffusivity) can be formed as:
( 4.28 )
where is a model constant. The transport equation from turbulent kinetic energy and
dissipation rate are modelled as (CFX Solver Theory Guide, ANSYS Inc, 2012):
*(
)
+
( 4.29 )
*(
)
+
( 4.30 )
The model constants are The middle term on the right hand side is the production of turbulence as a result of viscous
effects and is modelled as:
(
)
(
)
( 4.31 )
The model is an elegant and simple turbulence model but there are weaknesses. One
example is the over prediction of the spreading rate of the round jet. This can be fixed by
tweaking the model constants and/or add extra terms. The drawback is that when trying to fix
the model for a particular class of flows, the performance for other classes are inferior to the
original model formulation (Pope, 2000). Another problem with the model is that when
approaching the wall, the term becomes singular since goes to zero the wall
(Johansson & Wallin, 2012). For high Reynolds numbers, this is no problem since the log-law
is applicable for 30 < <500 and the velocity, turbulent kinetic energy and dissipation rate
can be modelled with wall functions. For low Reynolds numbers, the log-law does not apply
30
and modifications must be made to the transport equations (Versteeg & Malalasekera, 2007).
These modified models often require a y+
< 0.2 which is very costly and impractical (CFX
Solver Theory Guide, ANSYS Inc, 2012).
4.3.4.2 The k-ω model
This subchapter is based on CFX Solver Theory Guide, ANSYS Inc (2012). There are
naturally other ways to form the turbulent diffusivity. One can regard the variable which is
related to the turbulent viscosity as:
( 4.32 )
From Equations ( 4.28 )and ( 4.32 ), the relation between and can be seen to be:
( 4.33 )
can be seen as the frequency of the large scale eddies (Johansson & Wallin, 2012).
A basic turbulence model based on the transport of and was developed by Wilcox:
( )
*(
)
+ ( 4.34 )
( )
*(
)
+
( 4.35 )
The constants are
.
The value
will approach infinity when going close to wall since approaches zero but
it’s been seen that this can be avoided by setting the value of high enough (Johansson &
Wallin, 2012). Thus the k-ω model is better than the k-ε close to the wall. However, in the
form developed by Wilcox, the model is too sensitive the conditions of the free stream.
Menter developed a hybrid model which uses the k-ε model at the free stream and blends to a
k-ω formulation close to the wall, this is known as the Baseline k-ω model. Later on, an
addendum by the same author models the effect of the transport of the turbulent shear stress.
This model, known as the Shear Stress Transport (SST), has a very good prediction of flow
separation. This model is a popular one and is the turbulence model that is used in this thesis
4.3.5 Reynolds Stress Models (RSM) and Large Eddy Simulations (LES)
This subchapter is based on the work of Versteeg & Malalasekera (2007). The two equation
models presented previously relies on the Boussinesq assumption. However, flows with large
changes in mean strain rate or streamline curvature are not well captured by untweaked
turbulence models. Another approach is to calculate the transport equations for each Reynolds
stress component. This yields 6 additional partial differential equations (PDEs). Additionally,
a transport equation for the dissipation rate ε is needed. This approach is general and yields
better results for asymmetric channel flow, non-circular duct flow and also flows with
curvature. However, it is computationally expensive since it solves 7 additional PDEs. It
shares the dissipation rate modelling with the standard k-ε model and shows the same
deficiencies in certain flow cases.
31
Large Eddy Simulations (LES) is a rather different approach to the modelling of turbulence.
Large turbulent eddies can be seen as rather anisotropic and related to the geometry of the
problem whereas the small eddies are more isotropic. Therefore a single model that covers all
kinds of flow geometries are quite a challenge to find. LES acknowledges this problem and
sets out to calculate the structure of the time-dependent larger eddies but models the smaller
eddies which are assumed to be isotropic and thus simple to model. LES is rather expensive as
compared to the two-equation models but is has been observed that it might be sufficient with
twice the computational power as compared with RSM.
4.4 Numerical Schemes
To solve the transport equations presented previously, a numerical approach must be adopted
since there exists no analytical solution for the full Navier-Stokes equations. This subchapter
is based on the CFX Solver Theory Guide, ANSYS Inc (2012) since it is the flow solver used
in this thesis.
CFX uses a finite volume approach where nodes are used to store variables and a control
volume is built up around the node. Furthermore, the governing equations are integrated over
the control volumes and gradient and divergence operators are converted to surface integrals
through Gauss’ divergence theorem. The volume and surface integrals are then discretized,
the details are not discussed here. Several of the terms in the now discretized equation must be
evaluated not at the node where all the variables are stored but rather at the surface of the
control volume, denoted as integration point, ip. CFX uses shape functions to approximate the
values.
The advective term for a quantity at the integration points can be calculated in several
ways. A general form can be seen in Equation ( 4.36 ) where different choices of β and
yield different schemes. The vector from the upwind node to the integration point is given by
.
( 4.36 )
If β is equal to zero, a first order upwind scheme is obtained. While robust, it will introduce
diffusive errors which will smear out the solution. If β is between 0 and 1 and is taken as
the “average of the adjacent nodal gradients”, the numerically induced diffusivity will be
reduced. For β of 1, the scheme is formally second order in space accurate, however
dispersive errors might appear which manifest as wiggles. The High Resolution Scheme
varies β for each node and tries to keep it as close to 1 as possible and is taken from the
upwind node only. This is the advective scheme used in this thesis. Diffusion terms and the
pressure gradient term are evaluated through a standard finite element treatment and the
details will not be treated here. The control volumes are co-located for all quantities, this is
known to induce a decoupled pressure field. However, a different discretization (not treated
here) is implemented to remedy this. The temporal discretization is given by the backward
Euler method. The First Order Backward Euler is fully implicit, bounded, robust and there is
no limit in step size, however the first order accuracy will make it prone to diffusive
discretization errors which smooth sharp temporal gradients. A higher order discretization is
the Second Order Backward Euler which is also implicit, robust and has no step size limit but
32
may yield numerical oscillations. CFX solves the unsteady equations even for steady state
problems where it approaches a steady solution through advancing the approximate solution
in time.
CFX is a coupled solver which solves for u and p as an equation system and is fully implicit.
The coupled solver is more robust, general, simple and efficient. However, it requires large
storage space. Segregated solvers use an approximate pressure to solve the momentum
equations and then correct the pressure field. This results in a larger number of iterations as
compared with the coupled solver. However, segregated solvers do not need to store as much
information as coupled solvers.
33
5 Methodology
5.1 Parameters of study
Binder et al. (2014) investigated experimentally the effect of L/D, w/D, and pressure ratio on
the discharge coefficient and produced a correlation valid for:
( 5.1 )
The goal with the current study is to supplement the work with data from radiused orifices
with and without cross flow. Binder et al. (2014) used 1704 data points for the creation of the
correlation whereas this study will use around 110 data points due to time constraints
combined with the fact that a CFD simulation for one data point takes longer time than its
experimental counterpart. The parameters studied and their ranges were:
( 5.2 )
This work is divided into two parts. The first part, Study I, concerns flow through orifices
without inlet cross flow and the second part, Study II, treats flow through orifices with inlet
cross flow. The orifices studied for both parts had a diameter D of 7 mm and their dimensions
are presented in Table 5.1 and Table 5.2.
Table 5.1: Orifice geometry, Study I.
L/D [-] (r/D)1 [-] (r/D)2 [-] (r/D)3 [-] (r/D)4 [-] (r/D)5 [-] (r/D)6 [-]
0.4 0 0.04 0.08 0.16 0.20 -
1.5 0 - - - - -
2.0 0 0.04 0.08 0.16 0.20 0.30
4.0 0 0.04 0.08 0.16 0.20 0.30
34
Table 5.2: Orifice geometry, Study II.
L/D [-] (r/D)1 [-] (r/D)3 [-] (r/D)4 [-] (r/D)6 [-]
2.0 0 0.08 0.16 0.30
4.0 0 0.08 0.16 0.30
The orifice with L/D = 1.5 in Study I was used for verification and validation purposes only.
It can also be seen that for L/D = 0.4, the largest radius to diameter ratio was not applied due
to the fact that such a large radius is rarely used in industrial application for the specific L/D
and was thus disregarded in this work.
Due to time constraints Study II contains fewer geometries than Study I, it can be seen for
example that no orifices with L/D = 0.4 were included.
5.2 Geometry
This work is a continuation of the experimental study by Binder et al. (2014) and great effort
was taken in ensuring that the computational domain would accurately represent the
experimental setup. A schematic diagram of the test rig used by Binder et al. (2014) can be
seen in Figure 5.1.
Figure 5.1: Schematic diagram of test rig (Binder et al., 2014).
A pressurized air system was connected to the flow control valve from which the flow was
regulated. The Coriolis flow meter measured the mass flow of the passing air. The air then
entered into a pressure vessel and finally emptied out into the ambient environment through
an orifice due to the pressure difference. The orifice diameter D was 7 mm the diameter of the
pressure vessel was 330.2 mm or approximately 47D. To capture the same flow physics in the
CFD calculation, the geometry modelled included the pressure vessel, the orifice and the
ambient environment. It is however infeasible to model the whole pressure vessel due to the
sheer amount of computational cells needed so a simplification is needed. Also, a challenge
arises in modelling the ambient environment. How should it be modelled so physically
meaningful boundary conditions can be applied?
Binder et al. (2014) determined the pressure vessel to be large enough that the dynamic to
total pressure ratio did not exceed 0.0185%. The flow in the vessel can be considered to be
35
quiescent in large parts of the pressure vessel, save the region closest to the orifice. The
modelling of the pressure vessel can then be restricted to the region closest to the orifice.
From the orifice’s point of view, the flow approaches from all directions as the diameter of
the pressure vessel is much larger than the diameter of the orifice. To model this, a
hemisphere can be used to represent the pressure vessel since a boundary condition with a
normal velocity direction can be set on the curved hemisphere surface. The orifice is
geometrically a cylinder and modelled as such. The ambient environment was also modelled
as a hemisphere. From the orifice exit, the radial distance to the hemisphere boundary is equal
for all points on the boundary: Therefore a boundary condition on the curved surface of the
hemisphere applies equally on the orifice and is sufficient for modelling the ambient
environment.
The hemisphere used for modelling the pressure vessel can be seen to the left in Figure 5.2 (a)
and is much smaller than the hemisphere used for modelling the ambient environment which
can be seen to the right in the same figure. The reason for the large hemisphere representing
the ambient environment is that the jet emerging into the ambient environment can potentially
cause convergence problems if the gradients are still strong at the boundary, therefore a larger
geometry was used to reduce the risk of large gradients at the boundary.
(a) (b)
Figure 5.2: Geometry of computational domain L/D = 4.0, r/D = 0.08. (a) Study I. (b) Study II.
For Study II, where cross-flow is imposed on the inlet, the geometries from Study I could be
used with a slight modification. To account for cross-flow, an inlet box has to be attached to
the inlet hemisphere. The situation is now quite different for the pressure vessel and it can be
seen as a duct with an orifice in the side of the duct. An example of the geometry for Study II
can be seen in the Figure 5.2 (b). The hemisphere representing the pressure vessel is
henceforth denoted as Hemisphere I, the hemisphere representing the ambient environment as
Hemisphere II and the cube attached to Hemisphere I as Cross-flow cube. The dimensions can
be found in Table 5.3.
Table 5.3: Dimensions of parts in the computational domain.
L/D = 0.4 (r/D)1 [-] (r/D)2 [-] (r/D)3 [-] (r/D)4 [-] (r/D)5 [-] (r/D)6 [-]
Hemisphere
I diameter 14D 10D 10D 10D 10D -
Hemisphere
II diameter 20D 20D 20D 20D 20D -
Cross-flow
cube length - - - - - -
L/D = 1.5 (r/D)1 [-] (r/D)2 [-] (r/D)3 [-] (r/D)4 [-] (r/D)5 [-] (r/D)6 [-]
36
Hemisphere
I diameter
14D - - - - -
Hemisphere
II diameter 20D - - - - -
Cross-flow
cube length - - - - - -
L/D = 2.0 (r/D)1 [-] (r/D)2 [-] (r/D)3 [-] (r/D)4 [-] (r/D)5 [-] (r/D)6 [-]
Hemisphere
I diameter 14D 10D 10D 10D 10D 10D
Hemisphere
II diameter 20D 20D 20D 20D 20D 20D
Cross-flow
cube length 12D (8D) - 12D - 12D 12D
L/D = 4.0 (r/D)1 [-] (r/D)2 [-] (r/D)3 [-] (r/D)4 [-] (r/D)5 [-] (r/D)6 [-]
Hemisphere
I diameter 14D 10D 10D 10D 10D 10D
Hemisphere
II diameter 20D 20D 20D 20D 20D 20D
Cross-flow
cube length 12D - 12D - 12D 12D
For sharp-edged orifices, (r/D)1, the diameter of Hemisphere I can be seen to be larger than
for other geometries, this is due to the fact that it was discovered later on that reducing the
diameter of Hemisphere 1 from 14D to 10D did not alter the mass flow significantly.
Subsequent cases were done with a reduced diameter to save computational time. In the case
with inlet cross-flow, the Cross-flow cube length is 12D for all geometries, however a
validation study was made with a length of 8D instead for L/D = 2.0, (r/D)1 in order to better
replicate the experimental study.
5.3 Computational mesh
An adequate computational mesh is crucial when performing CFD calculations and a large
portion of the time of the work was spent on the creation of the computational meshes. A
mesh was made for each geometry in ICEM 14.5.7, however they are quite similar and the
basic structure of the mesh can be divided into three categories. A detail of the mesh element
counts can be found in Appendix A.
5.3.1 Mesh type I, sharp-edged orifices
For all sharp-edged orifices, i.e. r/D = 0, the final mesh is composed of three different meshes
connected to each other with a General Grid Interface (GGI). The inner structure was meshed
with a structured hexahedral mesh and was composed of a cube connected to an orifice which
in turn was connected to a second cube. The reason for this strategy was that a structured
hexahedral mesh was desired since it offers more control when meshing. However this was
not realizable for the full geometry and the hemispheres were meshed with unstructured
tetrahedrals.
37
(a) (b)
Figure 5.3: L/D = 4.0, r/D = 0. (a) Structured hexahedral mesh. (b) Unstructured tetrahedral mesh, Hemisphere I.
(a) (b)
Figure 5.4: L/D = 4.0, r/D = 0. (a) Unstructured tetrahedral mesh, Hemisphere II. (b) Final mesh.
5.3.2 Mesh type II, radiused orifices
The structured mesh in radiused orifices could, due the nature of the geometry, be extended to
include Hemisphere I. This resulted in a structured hexahedral mesh composed of Hemisphere
I, the orifice and a cube, see Figure 5.5 (a) , and an unstructured tetrahedral mesh identical to
the type I mesh seen in Figure 5.4 (a). The meshes were also connected through a GGI and
the final form can be observed in Figure 5.5 (b).
(a) (b)
Figure 5.5: L/D = 4.0, r/D = 0.08. (a) Structured hexahedral mesh. (b) Final mesh.
38
5.3.3 Mesh type III, cross-flow
To accommodate for inlet cross-flow, a cube had to be attached to the Hemisphere I. For
sharp-edged orifices (mesh type I), the structured mesh seen in Figure 5.3 (a) had already the
shape of a cube so the only modification was to enlarge the cube seen to left in the same
figure and disregard the Hemisphere I mesh seen in Figure 5.3 (b). An example of the
structured hexahedral mesh for sharp-edged orifices adapted for cross-flow can be seen in
Figure 5.6 (a), also the final mesh can be seen in Figure 5.6 (b).
(a) (b)
Figure 5.6: L/D = 4.0, r/D = 0, cross-flow. (a) Structured hexahedral mesh adapted for cross-flow. (b) Final mesh.
For radiused orifices (mesh type II), an unstructured tetrahedral mesh was applied to a cube
geometry and then connected to Hemisphere I through a GGI. The unstructured tetrahedral
mesh can be seen in Figure 5.7
Figure 5.7: L/D = 4.0, r/D = 0.08, cross-flow. Unstructured tetrahedral mesh, Cross-flow cube.
5.4 General Parameters
The commercial CFD solver ANSYS CFX 14.5 was used. A RANS-based turbulence model,
k-ω SST was used to account for turbulent flow. The fluid was air treated as an ideal gas to
account for compressibility. The total energy equation was used since the flow regime is in
most cases above the compressible limit (M > 0.3). The viscous work term was disregarded
since it is “negligible for most flows” and the turbulent Prandtl number, which states the
relation between the turbulent viscosity and the turbulent diffusivity , was set to its
default value of 0.9 (CFX-Solver Theory Guide, ANSYS Inc, 2012).
The advection scheme was set to “High resolution”. This option utilizes a blend factor and the
formal order of accuracy is aimed to be as close to second order as possible. Similarly, the
turbulent scheme was also set to “High resolution” which uses the same advection scheme.
39
To treat the turbulent boundary layer an automatic wall function was used. It switches from a
low Reynolds formulation for < 2 to a wall function for higher values of (CFX Solver
Modeling Guide, ANSYS Inc, 2012). Due to the separation in the orifice, it was believed that
a y+
≈ 1 should hold at the orifice wall to yield more accurate results and the meshes was
created accordingly. In the regions where no flow separation was expected, the meshes were
created with the criterion of y+
≤ 30. For certain cross-flow simulations, it was revealed that y+
= 3 at the inlet region in the orifice but this issue was not addressed due to time constraints.
Steady simulations were aimed for but for certain flow cases with short orifices (L/D = 0.4),
convergence could not be reached and oscillating residuals were present which hints an
unsteady behaviour. Transient simulations were performed and the desired quantities were
time averaged.
The convergence criterion was set as the RMS values of the residual to be less than
for steady state simulations and for transient simulations. Two additional tools were
used to determine if convergence has been achieved. Firstly, monitor points were used in
different regions of the computational domain to examine if the quantities had become steady
(or oscillating with finite and consistent amplitude for transient simulations). The quantities
monitored were velocity, pressure and turbulent kinetic energy. Secondly, the domain
imbalance of mass, momentum and energy were taken into consideration. The highest
imbalance was found to be approximately 0.3 % for the energy, this can be taken as
acceptable.
A summary of the most important solver settings can be seen in Table 5.4.
Table 5.4: Solver settings.
Fluid Air ideal gas
Turbulence model K-ω SST (beta reattachment
modification for L/D > 0.4)
Wall function Automatic
Heat transfer Total energy (viscous work disregarded)
Turbulent Prandtl number 0.9
Transient scheme Second order backward Euler
Advection scheme High resolution
Turbulent scheme High resolution
Convergence criterion RMS Steady: 1∙10
-5
Transient: 1∙10-4
5.5 Boundary conditions
The boundary conditions will differ between the cases with and without cross-flow and will
be treated in the following subchapters.
5.5.1 No cross-flow
For flow cases without cross-flow, the boundary conditions should represent the experimental
study by Binder et al. (2014). This calls for a pressure inlet/pressure outlet boundary
condition. Hemisphere I, which represents the pressure vessel, had a boundary condition of
type “Inlet” on the curved surface with a total pressure equal to the pressure in the vessel
specified with velocity entering normally through the boundary. The total pressure here is
equivalent to the static pressure due to the quiescent air which is also the situation in the
40
pressure vessel. Hemisphere II, representing the ambient environment, had a boundary
condition of type “Opening” with a specified opening pressure equal to the ambient pressure.
The “Opening” boundary condition allows the flow to re-enter the domain if required (CFX
Solver Modeling Guide, ANSYS Inc, 2012). The boundary condition “Outlet”, which does
not allow flow to re-enter the domain was also evaluated but no convergence could be reached
and backflow was present. As a consequence the decision was made to use the “Opening”
type. The walls were modelled as smooth adiabatic walls with a no-slip condition. The
pressure vessel was large and the flow velocity is near zero in it, thus the turbulence intensity
level Tu was set to 1%.
CFX offers the possibility of having two different meshes connected to each other through a
General Grid Interface (GGI). They do not have to match each other and can even be of
different types. All the connections between the different meshes explained in subchapters
5.3.1-5.3.3 are made by a GGI. A schematic of the boundary conditions can be seen in Figure
5.8. Note that for sharp-edged orifices, an additional GGI is inserted between Hemisphere I
and the cube.
Figure 5.8: Boundary conditions. L/D = 4.0, r/D = 0.08.
The boundary condition values were taken from experimental data and varied from case to
case. A reference pressure was set in the solver and the pressures defined on the boundaries
are gauge pressures. A summary of the boundary conditions for this subchapter can be found
in Table 5.5.
Table 5.5: Boundary conditions, no cross-flow.
Reference pressure ≈100000 Pa
Inlet
Total pressure (gauge)
Total temperature
Turbulence intensity
Flow direction
≈1000-70000 Pa
≈292-296 K
1%
Normal to boundary condition
41
Outlet
Opening pressure (gauge)
Opening temperature
Turbulence intensity
Flow direction
0 Pa
292.15-293.15 K
1%
Normal to boundary condition
Wall
Mass and momentum
Heat transfer
Wall roughness
No slip
Adiabatic
Smooth
GGI
Boundary type
Mass and momentum
Turbulence
Heat transfer
Interface
Conservative interface flux
Conservative interface flux
Conservative interface flux
5.5.2 Cross-flow
The boundary conditions changes when studying cross-flow. A cross-flow velocity has to be
imposed on the inlet and a supplementary outlet has to be defined. A schematic of the
boundary conditions can be seen for L/D = 4.0, r/D = 0.08 in Figure 5.9.
Figure 5.9: Boundary conditions, cross-flow.
The pressure ratio across the orifice was desired to be 1.1 which corresponds to engine-like
conditions. This was obtained by setting a static pressure condition on the outlet to 10132.5 Pa
(g). The inlet boundary condition was set as normal speed. The author believes that total
pressure could have been used as well. The study of Binder et al. (2014) did not include cross-
flow so the reference pressure, total temperature at the inlet and opening temperature are set
as constant values for all cases. The remaining boundaries (walls and GGI) were identical to
the ones without cross-flow. A summary of the boundary conditions that have been altered is
presented in Table 5.6.
42
Table 5.6: Boundary conditions, cross-flow.
Reference pressure 101325 Pa
Inlet
Normal speed
Total temperature
Turbulence intensity
22.2-150.0 m/s
294 K
1 %
Outlet
Average static pressure (gauge)
10132.5 Pa
Opening
Opening pressure (gauge)
Opening temperature
Turbulence intensity
0 Pa
293.15 K
1 %
5.6 Mesh verification
A computational mesh should ideally be able to accurately model the flow situation as cost-
efficiently as possible. It is therefore necessary to verify that the mesh used is not too coarse
for the application, since it will yield inaccurate results but also not too fine since the
additional computational effort will not yield any different result. Due to time constraints,
mesh verification was only performed for flow cases without cross-flow. Preferably, mesh
verification should have been carried out for flow cases with cross-flow since it represents
different flow physics.
5.6.1 No cross-flow
Two geometries were chosen with each representing its mesh type.
5.6.1.1 L/D = 1.5, r/D = 0
A mesh verification was performed on a sharp-edged orifice with L/D = 1.5, the meshes are of
type I seen in subchapter 5.3.1. Two different pressure ratios, 1.1 and 1.5, each representing
an extreme of the range of pressure ratios, were investigated. The main variable considered is
the mass flow since it has a crucial part in determining the discharge coefficient.
Four different meshes, numbered 1-4 were generated. All meshes are based on Mesh 3 and
scaled up or down. The scaling was done in two different ways. For the structured hexahedral
mesh, a refinement factor was specified. A refinement factor less than one is equivalent with a
coarsening of the mesh, whereas a refinement factor larger than one corresponds to a
refinement of the mesh. However, the refinement of the unstructured tetrahedral mesh was
performed through a scale factor. A scale factor less than one yields a refinement of the mesh,
whereas a scale factor larger than one results in a coarsening of the mesh. It might also be
worth to note that while the mesh was scaled the initial spacing of the first element was held
constant to retain the same desired value of y+ for the structured hexahedral mesh. A summary
of the details can be seen in Table 5.7.
43
Table 5.7: Mesh verification. L/D = 1.5, r/D = 0, mesh parameters.
Mesh Elements
Elements
across the
orifice
Refinement
factor
hexahedral
mesh
Global
element
scale factor
tetrahedral
mesh
1 1452755 87 0.6 1.4
2 3249319 115 0.8 1.2
3 5940630 140 1 1
4 10551087 168 1.2 0.8
For each mesh, the refinement factor and scale factor does not correspond one-to-one. An
estimation of how consistent the refining and coarsening is between hexahedral and
tetrahedral elements is given by taking the average volume of an element and compare it with
the average volume of the element in Mesh 3. From this, it can be seen in Table 5.8 that the
refinement/coarsening gives a consistent ratio between both element types except for Mesh 1
where it can be seen that the element size ratio for the hexahedral elements is roughly 60%
larger than the element size ratio for the tetrahedral elements.
Table 5.8: Average element volume comparison normalized with the element volume of Mesh 3.
Mesh Avg. hexahedral volume
ratio [-]
Avg. tetrahedral volume
ratio [-]
1 4.3690 2.7425
2 1.8439 1.7184
3 1 1
4 0.5699 0.5085
The results from the mesh verification for a pressure ratio p1/p2 = 1.1 are presented in Table
5.9. Each result is compared with the finest mesh, Mesh 4.
Table 5.9: Mesh verification. L/D = 1.5, r/D = 0, results.
Mesh Pressure
ratio [-]
Mass flow
[g/s]
Relative
difference
with mesh 4
[%]
Separation
region
length [mm]
Relative
difference
with
mesh 4
[%]
1 1.1 4.602 0.18 5.49 6.3
2 1.1 4.600 0.13 5.65 3.58
3 1.1 4.596 0.05 5.86 0
4 1.1 4.594 - 5.86 -
It can be seen that the difference in mass flow is very low, with the highest difference of
0.18% for the coarsest mesh. However, the separation region lengths in the orifice (extracted
by sampling along the orifice length for all meshes to enable comparison) vary noticeably
with the largest difference being 6.3% for the coarsest mesh. However, the mass flow is the
main quantity of concern so although the separation region lengths differ, Mesh 1 is sufficient
for future studies. Furthermore, the separation region for Mesh 1 and Mesh 4 is consistent in
shape which can be seen in Figure 5.10 (a) and (b) by plotting the negative axial velocities.
44
(a) (b)
Figure 5.10: Separation region illustrated by contour plots of negative axial velocity. (a) Mesh 1. (b) Mesh 4.
It is also important to ensure that the profiles of certain quantities are consistent between the
meshes. In this verification, the axial velocity profiles and turbulent kinetic energy (TKE)
profiles were taken from lines across the orifice as illustrated in Figure 5.11 for the four
meshes. Only the closest and farthest lines will be treated.
Figure 5.11: Lines across the orifice.
The values are given in a non-dimensionalized form and are based on the Cartesian coordinate
system:
The axial distance is denoted as x/L where L is the length of the orifice.
The distance across the orifice, in this case y and z, are translated and normalized with
the orifice diameter D and denoted as y/D and z/D respectively
Furthermore, the axial velocity and turbulent kinetic energy are normalized with the
largest values obtained in Mesh 4.
Firstly, the closest line to the orifice inlet at x/L=0.095, is examined. It can be observed in
Figure 5.12 that the axial velocity profiles are similar in shape and in magnitude; however
discrepancies can be observed at the extremes where the profiles do not match well. Figure
5.12 also offers an insight of the orifice flow field. There is a negative velocity region close to
the wall due to the separation region. The axial velocity then increases drastically when
moving further away from the wall, this is due to flow acceleration due to the nozzle formed
by the separated region. The free stream velocity is seen at y/D = 0.5, i.e. in the middle of the
orifice, to be around 80% of the maximum velocity.
45
It is observed in Figure 5.13 that the relative difference in turbulent kinetic energy between
Mesh 1 and Mesh 4, i.e. the coarsest and the finest mesh is quite elevated at the separation
region, a comprehensive explanation of this phenomenon cannot be presented but it is evident
that mesh size affect the turbulent kinetic energy. Also, the turbulent kinetic energy can be
observed to be largest close to the wall and rather low in the free stream.
Figure 5.12: Axial velocity profile across the orifice at x/L = 0.095.
Figure 5.13: Turbulent kinetic energy profile across the orifice at x/L = 0.095.
46
Figure 5.14: Axial velocity profile across the orifice at x/L = 0.952.
Figure 5.15: Turbulent kinetic energy profile across the orifice at x/L = 0.952.
Secondly, it is interesting to look at a line outside the separation region. The farthest line from
the orifice inlet is located at x/L = 0.952. The axial velocity profiles can be regarded in Figure
5.14 it is concluded that the velocity profiles are virtually identical to each other. Furthermore,
it can be noted that the turbulent velocity profiles in Figure 5.15 are quite similar to each other
to a larger extent than in the separation region seen in Figure 5.13. The axial velocity and
47
TKE profiles were also evaluated for x/L = 0.476 but due to similarity with the profiles at x/L
= 0.952, the analysis is omitted and the figures can be found in Appendix B. In addition to
visually inspecting the axial velocity and TKE profiles, the relative difference of the
quantities between Mesh 1 and Mesh 4 can be extracted and is presented in Table 5.10 for x/L
= 0.095. The data was extracted through interpolation to enable the comparison.
Table 5.10: Relative difference between Mesh 1 and Mesh 4 in axial velocity and TKE at x/L = 0.095.
Location y/D [-] Rel. difference
axial velocity
[%]
Rel. difference
TKE [%]
Shear layer 0.0142 6.95 64.57
Centre line 0.5 0.87 15.35
It can be concluded that the relative differences are drastically high, especially at the shear
layer. Thus, for these quantities, it can be seen that the coarsest mesh, Mesh 1, does not
produce the same results as the finest mesh, Mesh 4. However, the important quantity here is
the mass flow and it can be concluded from the observations made in this subchapter that
although the axial velocity profiles and TKE profiles do not match exactly, they have a
similar shape. This was also observed for the shape of the separation region. Lastly, to ensure
that the flow is symmetric in the orifice, the axial velocity is plotted for two perpendicular
lines across the orifice for Mesh 1 at x/L = 0.095. Visual inspection of Figure 5.16 shows that
the axial velocities are indeed identical on the two perpendicular lines.
Figure 5.16: Axial velocity profile across the orifice at x/L = 0.095, Mesh 1.
From the analysis presented it can then be concluded that Mesh 1, i.e. the coarsest mesh, gives
a good performance for the task at hand and its parameters (growth rate, max size) were
carried over to the mesh of other sharp-edged geometries.
48
From this small study, it was seen that the initial mesh created, Mesh 3 was too fine for the
task and the same results could be achieved with a mesh with four times less elements.
A mesh verification study was also run for a pressure ratio of 1.5. However, for this pressure
ratio, there was evidence of supersonic regions in the orifice as seen in Figure 5.17. The
contour plot shows the Mach number and it can be seen that the Mach number exceeds unity
at the orifice inlet. Transitions from supersonic velocities to subsonic velocities involve shock
waves which require extensive work with the mesh to capture the shock properly. Moreover,
the solver might not reach converge as easily as in subsonic flow cases (CFX Solver
Modeling Guide, ANSYS Inc, 2012). Thus, the mesh verification for a pressure ratio of 1.5
cannot be taken as valid even though convergence was reached for all four meshes. It gives
however an estimate of where supersonic flow starts to appear. Based on this, the maximum
pressure ratio was limited to 1.4 for further studies. However, one validation case for a short
orifice with L/D = 0.4 had a pressure ratio up to 1.7 due to the fact the shorter orifices does
not choke as early as longer orifices as discussed in subchapter 3.3, and thus supersonic
regions are likely to occur later.
Figure 5.17: Contour plot of Mach number for Mesh 4, pressure ratio 1.5
5.6.1.2 L/D = 0.4, r/D = 0.04
The second mesh verification was performed for a short orifice with small inlet radiusing.
This mesh represents a Type II mesh as seen in subchapter 5.3.2. Only one pressure ratio
which was p1/p2 = 1.2 was utilised which is slightly higher than in the previous subchapter.
Three different meshes, numbered 5-7, were created in the same fashion as for the previous
mesh verification and are presented in Table 5.11.
Table 5.11: Mesh verification. L/D = 0.4, r/D = 0.04, mesh parameters.
Mesh Elements
Elements
across the
orifice
Refinement
factor
hexahedral
mesh
Global
element scale
factor
tetrahedral
mesh
5 1074824 107 0.8 1.2
6 1935475 130 1 1
7 3443258 156 1.2 0.8
49
The results from the mesh verification are presented in Table 5.12. The absence of the
separation region length column is explained by the fact that the flow does not reattach to the
orifice for the current flow case which can be seen in Figure 5.18.
Table 5.12: Mesh verification. L/D = 0.4, r/D = 0.04, results.
Mesh Pressure ratio [-] Mass flow [g/s]
Relative
difference with
mesh 7 [%]
5 1.2 6.54512 0.31
6 1.2 6.56538 0.0053
7 1.2 6.56573 -
Figure 5.18: Separation region illustrated by negative axial velocity contours. L/D = 0.4, r/D = 0.04, Mesh 5.
It can be deduced from Table 5.12 that Mesh 5, having three times fewer elements than Mesh
7 gives a mass flow that differs only 0.3 % from the latter mesh. However, as in the previous
case, the axial velocity and TKE profiles should be compared to ensure that they are
consistent before coming to a conclusion. The profiles were evaluated along three different
lines but only the line closest to the orifice inlet located at x/L = 0.4 will be presented here due
to similarity between profiles. The profiles at x/L = 0.6 and 0.8 can be found in Appendix B.
The axial velocity profiles in Figure 5.19 are virtually on top of each other even though it is
hinted at there exists a difference at the highest axial velocity.
The TKE profiles observed in Figure 5.20 are in fairly good agreement apart from the fact that
there exists a discrepancy between Mesh 5 and Mesh 7 at the peak values of TKE. This is
consistent with the observation seen in Figure 5.13, however the coarsest mesh here gives
lower values whereas the coarsest mesh in subchapter 5.6.1.1 gives higher values.
50
Figure 5.19: Axial velocity profile across the orifice at x/L = 0.4.
Figure 5.20: Turbulent kinetic energy profile across the orifice at x/L = 0.4.
The relative difference between Mesh 5 and Mesh 7 for axial velocity and TKE are, similarly
to the previous section, compared at the shear layer and the centre line for x/L = 0.4. The data
are sampled along the line to enable comparison between the two meshes which have
different amount of elements. Table 5.13 shows that there exist a difference between the
meshes but the difference is not as dramatic as the difference seen in Table 5.10.
51
Table 5.13: Relative difference between Mesh 5 and Mesh 7 in axial velocity and TKE at x/L = 0.4.
Location y/D [-] Rel. difference
axial velocity
[%]
Rel. difference
TKE [%]
Shear layer 0.01 1.21 4.76
Centre line 0.5 0.30 2.45
Finally, the axial velocity of two perpendicular lines at x/L = 0.4 in Mesh 5 are plotted in
Figure 5.21. The axial velocity profiles are identical but the sharp-eyed can observe that the
mesh is slightly skewed in the free stream since the location of the data points do not match
on two occasions. However, since the profiles otherwise are identical, this skewness is not
believed to affect the solution noticeably. From the analysis in this section, it can be
concluded that the coarsest mesh, Mesh 5, should be sufficient for the current task. Its
parameters (growth ratio, max size) were subsequently carried over to all radiused orifices.
Figure 5.21: Axial velocity profiles across the orifice at x/L = 0.4, Mesh 5.
5.6.2 Cross-flow
Due to time constraints, no mesh verification could be performed for cross-flow cases.
However, the meshes used are partly from the flow cases without cross-flow which were
verified in subchapter 5.6.1.
5.7 Solver validation
One of the questions that are always present in CFD calculations is: How do we know that the
calculations gives answers that are in agreement with reality, i.e. how much confidence can
we have in our CFD results? Validation is the process of comparing the results given by CFD
calculations with results obtained experimentally. In this section, validation will be made
against data from Binder (2013) for flow cases without cross-flow but also with data from
52
Hüning (2012) for flow cases with cross-flow. However, before validating the CFD
calculations, a small study was made to compare difference turbulence models with data from
Binder (2013).
5.7.1 Turbulence models
An overview of the flow cases and the turbulence models studied can be seen in Table 5.14
and Table 5.15. Note that this study was made for cases without cross-flow only. The model
of choice up to this point has been the k-ω SST with beta reattachment option but this study
was made to benchmark several turbulence models and compare them with experimental data
in order to decide which model best suits the needs.
Table 5.14: Geometrical and physical parameters for turbulence model study.
L/D [-] r/D [-] p1/p2 [-]
0.4 0 1.2
1.5 0 1.2
The two geometries were chosen since they represent two different flow regimes, the flow in
the short orifice will not reattach to the orifice wall whereas the flow in the long orifice will
reattach as already seen in the mesh verification study for L/D = 1.5.
Table 5.15: Overview of studied turbulence models.
Turbulence model Option 1 Option 2
k-ω SST
- -
k-ω SST
Reattachment
modification (beta) -
k-ω SST Reattachment
modification (beta) Curvature correction
Spalart-Allmaras (beta)
- -
RNG k-ε - -
The k-ω SST model is explained in section 4.3.4.2. The option “Reattachment modification”
is a beta function that is available in CFX only. It has been observed that while most RANS
model accurately predicts the separation point, they overpredict the length of the separation
region. This model seeks out to correct this by adding an extra production term for TKE at the
separated shear layer and the results are in better agreement with experiments compared to the
standard k-ω SST model. However, the effects of the model have been shown to vanish when
the grid is refined and it is therefore in a beta stage (Lechner et al., 2010). The inlet of the
orifice will experience separation and reattachment, therefore this option was included in the
study. The curvature correction option tries to correct the fact that two-equation models are
insensitive to streamline curvature by adding a modification to the TKE production term
(CFX-Solver Theory Guide, ANSYS Inc, 2012). This option was chosen since the flow at the
inlet of the orifice experience a curvature when entering the orifice. The Spalart-Allmaras
model is a one equation model also in beta stage in CFX 14.5, a general description of it can
be found in FLUENT Theory Guide 14.5 (ANSYS Inc, 2012). It is originally designed for
wall-bounded flows with an application in aerospace. Relatively large errors are expected for
certain shear flows and it is not calibrated for general industrial application. It was chosen
since it is a fairly common model used. The RNG k-ε model is a two equation model which
53
has minor changes from the standard k-ε model and the details are treated in CFX-Solver
Modeling Guide (ANSYS Inc, 2012).
From Table 5.16, it can be concluded that the Spalart-Allmaras model has quite good
performance for both cases. However, there exists a reason to why this model is not available
generally (even though the author has yet to find it) and it has its weaknesses as stated
previously. In the same table it can be seen that the standard k-ω SST model has the best
performance for a short orifice with L/D = 0.4. For a long orifice, L/D = 1.5, it can be
observed that the k-ω SST with reattachment modification and curvature correction performs
best. The RNG k-ε model did not converge for the case with L/D = 0.4 and is thus not
included in the comparison.
Table 5.16: Relative difference in mass flow with Binder (2013).
L/D = 0.4, r/D = 0, p1/p2 = 1.2
Turbulence model Relative difference of mass flow with
Binder (2013). [%]
k-ω SST
0.72
k-ω SST
reattachment modification (beta) 1.94
k-ω SST
reattachment modification (beta),
curvature correction
4.82
Spalart-Allmaras (beta) 0.70
L/D = 1.5, r/D = 0, p1/p2 = 1.2
Turbulence model Relative difference of mass flow with
Binder (2013) [%]
k-ω SST
2.35
k-ω SST
reattachment modification (beta) 1.84
k-ω SST
reattachment modification, curvature
correction
1.49
Spalart-Allmaras 1.85
The mesh verification study in section 5.6.1.1, i.e. L/D = 1.5, was performed prior to this
study with the k-ω SST with reattachment modification turbulence model. It is observed that
this option gives a slightly larger difference. For the sake of consistency, it was decided to
continue with this turbulence model for long orifices i.e. orifices with L/D > 0.4. On the other
hand, for orifices with L/D = 0.4, the decision was made that the turbulence model used
would be the standard k-ω SST model. The mesh verification in section 5.6.1.2 was done
after this study and thus the standard k-ω SST model was used there.
To summarize, for short orifices with L/D = 0.4, the k-ω SST model was used whereas for
longer orifices with L/D > 0.4, the k-ω SST model with reattachment option was used. It can
be discussed which model would perform better for short orifices, L/D = 0.4, with a large
radiusing since the flow will reattach. Would the reattachment modification give more
accurate results as it does with L/D = 1.5 which has reattachment? It is however impossible to
deduce since there is no data to validate against for this case.
54
5.7.2 Model validation
Three validation cases were made, each chosen to represent its problem class. The first case
was chosen since no reattachment is expected for short orifices, the second one since
reattachment was expected. Lastly, the third one represents the cross-flow cases. Due to lack
of data in open literature, all the validation cases were made for sharp-edged orifices whereas
the main focus of this study concerns radiused orifices. A summary can be seen in Table 5.17.
Table 5.17: Summary of validation cases.
Case L/D r/D Variable Validation
source
1 0.4 0 ϵ
[1.01;1.7]
Binder
(2013)
2 1.5 0 ϵ
[1.01;1.4]
Binder
(2013)
3 2.0 0 U/wax ϵ
[0.17;0.74]
Hüning
(2012)
Validation case 1 was chosen not only due to the fact that it concerns a short orifice, but also
due to the fact that Binder (2013) found an interesting behavior for this geometry. In the paper
by Binder et al. (2014) which is based on the work of Binder (2013), a phenomena is
discussed where the author noticed a sudden rise of the discharge coefficient usually together
with a high frequency sound for a range of pressure ratios between 1.1 to 1.75 for L/D
between 0.33 to 0.75. This trend was named bubble after its shape in the plots for the
discharge coefficient as a function of the pressure ratio. For L/D = 0.33 and L/D = 0.4, the
discharge coefficient returns to the same trend seen before the bubble when going beyond the
bubble region. It is theorized that reattachment is not the underlying mechanism since the
trend of the discharge coefficient would be altered outside the bubble region and not be the
same before and after the bubble. Binder et al. (2014) speculate that the high frequency sound
heard indicates that resonance might be present in the boundary layer in the orifice which
could generate to oscillations of the vena contracta size. This would then yield higher a
discharge coefficient since the time averaged size of the vena contracta would always be
larger than the cross-sectional area without oscillations.
The results for case 1 can be seen in Figure 5.22. Firstly, the CFD calculation underpredicts
the discharge coefficient slightly. The bubble phenomena can be seen for pressure ratios
between 1.3 to 1.65 where the linear increase of the discharge coefficient is not present as
seen below and above this range of pressure ratios. The CFD calculations do not capture the
bubble region well; the discharge coefficient can be seen to increase in a linear manner
without the slightest hint of assuming a bubble shape. Disregarding the region of the bubble,
the trend of the CFD data follows quite well the experimental trend with a largest relative
difference of 1.9% which strengthens the belief in the accuracy of future CFD calculations
where the flow does not reattach.
Before leaving this case, it might be interesting to look at the separation behavior inside the
orifice for a pressure ratio of 1.3 and 1.65 since the data is coherent between CFD and
experimental and surround the bubble region. In Figure 5.23 (a) and (b), contour plots of the
negative axial velocities are utilized to illustrate the separation region. From this, it can be
seen that there for both pressure ratios, the flow stays separated and does not reattach. Thus,
55
the statement of Binder et al. (2014) this bubble phenomena should not be a result of
reattachment can be seen to be at least correct for this geometry.
Figure 5.22: Validation case 1. L/D = 0.4, r/D = 0.
(a) (b)
Figure 5.23: Recirculation region illustrated by the contour plot of the negative axial velocities. (a) L/D = 0.4, r/D = 0, p1/p2= 1.3. (b) L/D = 0.4, r/D = 0, p1/p2= 1.65.
Validation case 2, with L/D = 1.5, r/D = 0 was also compared with experimental data from
Binder (2013), and the results are presented in Figure 5.24. It can be seen that trend between
CFD data and experimental data agrees quite well with CFD calculations again
underpredicting the discharge coefficient. The largest relative difference was found to be 2.8
%. In the mesh verification in subchapter 5.6.1.1 , it could be seen that the flow is reattached
for this geometry.
56
From these first two validation cases, it can be concluded that the CFD calculations give quite
reasonable results for the discharge coefficient for flow cases without inlet cross-flow and
more confidence can be put into future CFD results.
Figure 5.24: Validation case 2. L/D = 1.5, r/D = 0.
Validation case 3 is based on an experimental study by Hüning (2012). Due to time
constraints, this validation was made with a limited amount of data points, as compared to
previous validation cases. Please note that discharge coefficient used is the total discharge
coefficient together with the cross-flow ratio based on the ideal velocity based on total
pressure ratio . The static pressure ratio across the orifice was set to 1.05 in order to
compare with Hüning (2012) whereas the main work regarding cross-flow has a slightly
higher ratio of 1.10. It can be seen in Figure 5.25 that the CFD calculations once again
underpredict the discharge coefficient. However, the trend agrees between experimental
studies and CFD calculations although there is a constant discrepancy of . For
higher values of , this yields a lower relative difference but for lower values of , the
relative difference can amount to 10%. The experimental study had a quite high uncertainty of
around 4% for this pressure ratio which can be seen in the same figure as error bars.
The conclusion of validation case 3 is that even though there is a systematic difference
between experimental data and CFD data, the discharge coefficient trends are coherent
between them. Thus future cross-flow studies through the use of CFD can be taken as valid.
However, a CFD simulation is as the name implies just a simulation and all data should be
validated experimentally before implementing into real-life applications.
57
Figure 5.25: Validation case 3, L/D = 2.0, r/D = 0, p1/p2.
58
6 Results – Study I, no cross-flow
In this chapter the results are presented for Study I and the effects of different parameters will
be examined and compared with literature. Comparison will also be made with the work of
Binder (2013) and with existing correlations. A summary of the cases presented in the study
can be found in Table 6.1.
Table 6.1: Summary of flow cases in Study I.
L/D [-] r/D [-] p1/p2 [-]
0.4 0-0.20 1.05-1.4
2.0 0-0.30 1.05-1.4
4.0 0-0.30 1.05-1.4
Firstly, it can be observed that the highest pressure ratio has been limited to 1.4 due to the fact
that supersonic regions where found for a pressure ratio of 1.5 as seen in subchapter 5.6.1.1.
The lowest pressure ratio was set to 1.05 which is higher than the lowest pressure ratio in the
validation studies in subchapter 5.7.2. The Reynolds number, based on the orifice diameter D
and the isentropic velocity, is for all the flow cases larger than which yields a Reynolds
number independency as discussed in subchapter 3.2.3.2.
6.1 General overview
All the data are grouped after length to diameter ratio L/D and plotted as a function of
pressure ratio p1/p2 in Figure 6.1 (a) and (b), and in Figure 6.2 (a). The discharge coefficient is
shown to have a pressure dependency for all geometries to a certain degree. An augmentation
of the radius to diameter ratio r/D can be seen to increase the discharge coefficient up to a
certain limit where the rate of growth of the discharge coefficient declines. It can even be seen
that too large radiusing decreases the discharge coefficient for short orifices with a length to
diameter ratio L/D equal to 0.4. The effect of the length to diameter ratio L/D can be
examined by regarding Figure 6.2 (b). It is observed that an elongation of the orifice is
beneficial up the limit of L/D = 2.0, after which the discharge coefficient decreases. Finally,
the orifices with L/D = 4.0 is observed to have quite similar behaviour to orifices with L/D =
2.0 when comparing Figure 6.1 (b) and Figure 6.2 (a).
In following subchapters a detailed analysis, as opposed to the mere observations in the
current subchapter, will be carried out with focus on how the parameters and their interaction
affect the discharge coefficient.
59
(a) (b)
Figure 6.1: (a) Discharge coefficient CD as a function of pressure ratio p1/p2, L/D = 0.4. (b) Discharge coefficient CD as a function of pressure ratio p1/p2, L/D = 2.0.
(a) (b)
Figure 6.2: (a) Discharge coefficient CD as a function of pressure ratio p1/p2, L/D = 4.0. (b) Discharge coefficient CD as a function of length to diameter ratio L/D, r/D = 0.04.
60
6.2 Detailed analysis
Three variables that affect the discharge coefficient were studied and this subchapter seeks to
provide a detailed explanation of how the discharge coefficient depends on the variables both
individually and combined.
6.2.1 L/D and p1/p2
A representative radius to diameter ratio r/D = 0.04 is chosen. Figure 6.3 (a) shows that an
increase in pressure ratio p1/p2 increases the discharge coefficient for all length to diameter
ratios L/D. This is the result of the area increase of the vena contracta as separated region is
supressed which is in line with the discussion by Ward-Smith (1979). This can clearly be seen
by comparing the height of the separation region by measuring the cross orifice distance of
negative axial velocities. At the axial distance x/L = 0.5 for a length to diameter ratio L/D =
0.4 and radius to diameter ratio r/D = 0.04, the height is 0.033D for the case with pressure
ratio p1/p2 = 1.05, but 0.026D for the case with a pressure ratio of 1.4. It can also be observed
that the pressure ratio dependency is stronger for L/D = 0.4 compared with L/D = 2 and L/D =
4 which agrees with the findings of Deckker & Chang (1965). This might be illustrated by
Figure 6.4 (a) and (b) where it can be observed that the recirculation region covers the whole
orifice length for L/D = 0.4 and different pressure ratios presumably determines strongly the
shape of the recirculation region and thus the pressure recovery possible.
Secondly, it can be observed in Figure 6.3 (b) that the discharge coefficient CD increases as
the length to diameter ratio L/D increases, but up to a certain limit. The discharge coefficient
then decreases as L/D is increased. The increase of the discharge coefficient is due to the fact
that reattachment (and thus pressure recovery) is present for L/D = 2.0 but not for L/D = 0.4
(true for small radiusing) and can be seen in Figure 6.4. A decrease of the discharge
coefficient is seen between L/D = 2.0 and L/D = 4.0 and is presumed to be due to the
frictional losses in the orifice since the trend between these two L/D are similar as evidenced
in Figure 6.3 (a). For a pressure ratio p1/p2 = 1.4, the total pressure was mass flow averaged at
two locations, the first located behind the recirculation region at the axial distance x/L = 0.5,
and the second at the exit plane. The inlet conditions are included as reference. From Table
6.2, it is observed that the total pressure drop between the inlet and the region behind the
recirculation region is 2.6% for L/D = 2.0, r/D = 0.04 and 2.22% for L/D = 4.0, r/D = 0.04.
The pressure drop between the inlet and the exit plane is however 3.95% for the former case
and 5.31% for the latter case. It can then be presumed that the larger total pressure drop is due
to the frictional losses due to a longer orifice. This was also observed for the incompressible
discharge coefficient by Lichtarowicz et al. (1965).
Table 6.2: Pressure drop across the orifice.
L/D [-] r/D [-] Location Total pressure [Pa]
2.0
0.04
Inlet 142773
x/L = 0.5 139023
Exit plane 137138
4.0
Inlet 141782
x/L = 0.5 138638
Exit plane 135195
61
(a) (b)
Figure 6.3: (a) Discharge coefficient CD versus the pressure ratio p1/p2, r/D = 0.04. (b) Discharge coefficient CD versus the length to diameter ratio L/D, r/D = 0.04.
(a) (b)
Figure 6.4: Recirculation region illustrated by contour plots of the negative axial velocity. (a) L/D = 0.4, r/D = 0.04, p1/p2 = 1.4. (b) L/D = 2.0, r/D = 0.04, p1/p2 = 1.4.
6.2.2 L/D and r/D
A representative pressure ratio p1/p2 of 1.1 is chosen. It can be observed in the Figure 6.5 (a)
that an increased radiusing augments the discharge coefficient, but the beneficial effects starts
to decrease for high radius to diameter ratio r/D which agrees with the observations by Hay &
Spencer (1992) for chamfered orifices. It is seen that the discharge coefficient even decreases
when increasing r/D from 0.16 to 0.20 for L/D = 0.4. This can be explained by reattachment
seen in Figure 6.6. The flow almost reattaches for r/D = 0.16 which increases the through
flow. Conversely, the flow is separated at the exit of the orifice for r/D = 0.20 which
62
decreases the through flow. As a result of the large radiusing, the flow has a shorter distance
to reattach. This behaviour has been observed previously by Hay & Spencer (1992), Dittman,
et al. (2003) and Binder, et al. (2014) for chamfered orifices. The effective length as
proposed by Binder, et al. (2014) can thus be extended to radiused orifices as:
( 6.1 )
The effect of the length to diameter ratio L/D has been discussed in the previous subchapter
but an interesting effect can be noticed in Figure 6.5 (b), the discharge coefficient does not
decrease between the length to diameter ratio L/D = 2.0 and L/D = 4.0 for sharp edged
orifices with r/D = 0, the reason for this is unknown.
(a) (b)
Figure 6.5 (a) Discharge coefficient CD versus the radius to diameter ratio r/D, p1/p2 = 1.1. (b) Discharge coefficient CD versus the length to diameter ratio L/D, p1/p2 = 1.1.
(a) (b)
Figure 6.6: Recirculation region illustrated by contour plots of the negative axial velocities. (a) L/D = 0.4, r/D = 0.16, p1/p2 = 1.1. (b) L/D = 0.4, r/D = 0.20, p1/p2 = 1.1.
63
6.2.3 r/D and p1/p2
A representative L/D = 2.0 is chosen for this analysis. It can be noted that Figure 6.7 (a) is
identical to Figure 6.1 (b). Virtually no difference in the discharge coefficient can be found
between a radius to diameter ratio r/D = 0.16 compared with r/D = 0.30 as observed in Figure
6.7 (a). Observing the recirculation region for L/D = 2.0, r/D = 0.16 and p1/p2 = 1.4 in Figure
6.8, the size is small compared to the same flow conditions but for L/D = 2.0 and r/D = 0.04
seen in Figure 6.4 (b). Thus, the orifice is more similar to a convergent nozzle for r/D = 0.16
and increasing the radius to diameter ratio r/D only increases the discharge coefficient
marginally. Çengel & Cimbala (2006) defines a well-rounded inlet with negligible losses
having a radius to diameter ratio r/D = 0.20 which is in line with the current observations.
Figure 6.7 (b) reveals that the increase in pressure ratio p1/p2 only yields a slight raise in the
discharge coefficient for all radius to diameter ratios for this specific L/D.
(a) (b)
Figure 6.7 (a) Discharge coefficient CD versus the pressure ratio p1/p2, L/D = 2.0. (b) Discharge coefficient CD versus the radius over diameter ratio r/D, L/D = 2.0.
64
Figure 6.8: Recirculation region illustrated by a contour plot of negative axial velocities. L/D = 2.0, r/D = 0.16, p1/p2 = 1.4.
6.2.4 L/D, r/D and p1/p2
There are certain effects that are only present when the three parameters are varied. It has
been discussed previously that for short orifices, L/D = 0.4, excessive radiusing is
counterproductive. However, for high pressure ratios, the decrease in the discharge coefficient
is not as pronounced as for lower pressure ratios as observed in Figure 6.9. This can be
explained by once again regarding the separation regions. The recirculation regions for L/D =
0.4, p1/p2 = 1.4, r/D = 0.16 and r/D = 0.20 are shown in Figure 6.10 and it can be seen that
due to the high pressure ratio, the flow reattaches for both cases which enables the discharge
coefficient to remain fairly constant, something which is not the case in for low pressure
ratios as seen previously in Figure 6.6.
Figure 6.9: Discharge coefficient CD versus radius to diameter ratio r/D, L/D = 0.4.
65
(a) (b)
Figure 6.10: Recirculation region illustrated by contour plots of negative axial velocities. (a) L/D = 0.4, p1/p2 =1.4, r/D = 0.16. (b) L/D = 0.4, p1/p2 =1.4, r/D = 0.20.
The discharge coefficient for short orifices with a length of diameter L/D = 0.4 are usually
lower than for its longer counterparts. However, for high pressure ratios p1/p2 = 1.4 with a
generous radius to diameter ratio r/D = 0.16, the discharge coefficient is higher for L/D = 0.4
than for L/D = 2.0 and L/D = 4.0 as evidenced from Figure 6.11. This is due to the fact that
the flow is reattached even for L/D = 0.4 as seen in Figure 6.10 (a) and pressure recovery is
possible. Thus, longer orifices yield a slightly lower discharge coefficient due to frictional
losses.
Figure 6.11: Discharge coefficient CD versus pressure ratio p1/p2, r/D = 0.16.
6.2.5 Additional findings
Even though the pressure ratio p1/p2 was limited to 1.4 in order to avoid supersonic regions,
this was encountered for three flow cases listed in Table 6.3. However, the trends of the
discharge coefficients as seen in Figure 6.1 (b) and Figure 6.2 (a) are consistent and thus these
flow cases were included.
66
Table 6.3: Flow cases with supersonic regions.
L/D [-] r/D [-] p1/p2 [-]
2.0 0.04 1.4
2.0 0.08 1.4
4.0 0.08 1.4
The unsteady behaviour of short orifices have been mentioned previously and can be seen
clearly in Figure 6.12 where contour plots of the negative axial velocities for L/D = 0.4, r/D =
0.16, p1/p2 = 1.05 are shown. Figure 6.12 (a) shows the instantaneous axial velocity at 0.01s
and the Figure 6.12 (b) the time-averaged axial velocity. Visual inspection reveals that there is
a considerable size difference which indicates that a transient behaviour is present.
(a) (b)
Figure 6.12: Comparison of unsteadiness of the recirculation region, L/D = 0.4, r/D = 0.16, p1/p2 = 1.05. (a) Instantaneous axial velocity at 0.01 s. (b) Time-averaged axial velocity.
67
6.3 Comparison with chamfered orifices
Chamfered and radiused orifices both lead to smaller separation regions, but how do they
compare against each other? Since this study is a continuation of the work by Binder (2013),
the range of variables was intentionally kept similar to enable comparison. Binder (2013)
studied chamfered orifices for length to diameter ratios L/D of 0.4 and 2.14 which can directly
be compared with the current study.
Figure 6.13: Comparison between chamfered and radiused orifices, L/D = 0.4.
From Figure 6.13, where orifices with a length over diameter ratio of L/D = 0.4 are presented,
it can be observed that excessive chamfering leads to a lowered discharge coefficient as
discussed by Binder, et al. (2014). It is seen to occur already at w/D = 0.0276 whereas the
corresponding value for radiusing was found to be r/D = 0.16. This explains the higher
value recommended by Binder, et al. (2014). Moreover, it is apparent that small chamfering is
preferable over small radiusing. A chamfer to diameter ratio w/D = 0.022 yields the
approximately the same discharge coefficients as a radius to diameter ratio r/D = 0.04. Even
more striking, the discharge coefficient associated with a chamfer to diameter ratio w/D =
0.0276 is even higher than the discharge coefficient for a radiused orifice with a radius to
diameter ratio r/D = 0.08. Hay & Spencer (1992) found that for L/D = 0.25 and p1/p2 = 1.2,
chamfering was more beneficial than radiusing up to a chamfer to diameter ratio w/D = 0.08
where the radiused orifices yielded higher discharge coefficients. This was not found for cases
where L/D = 0.5 where chamfering perfomed slightly better than radiused orifices for all w/D
and r/D. The trend seen indicates that the orifices in Figure 6.13 behave similarly to the
orifices studied by Hay & Spencer (1992). Still, no definite conclusion can be made since the
largest chamfer to diameter studied by Binder (2013) was w/D = 0.0531.
In Figure 6.14 a comparison between chamfered (Binder, 2013) and radiused orifices is
presented for L/D = 2.0-2.14. It is observed that the adverse effect of excessive chamfering is
68
not present here which is also the case with radiused orifices as discussed previously. It can be
also observed that the benefits of increasing the chamfering are questionable for w/D >
0.0439, as higher chamfer to diameter ratios yield almost the same discharge coefficient. This
is also found to be the case for radius to diameter ratios r/D larger than 0.16 as discussed
previously. It is noticed that a radius to diameter ratio r/D = 0.08 yield a higher discharge
coefficient than w/D = 0.152.
Figure 6.14: Comparison between chamfered and radiused orifices, L/D = 2.0-2.14.
6.4 Comparison with correlations in literature
This subchapter will treat the comparison of the obtained data with existing correlations from
open literature. Three correlations are included: the Parker and Kercher correlation (Parker &
Kercher, 1991), the Idris and Pullen correlation (Idris & Pullen, 2005) and the Hüning
correlation (Hüning, 2008). In a review (Hüning, 2008), the Parker and Kercher correlation is
stated to be based on the incompressible discharge coefficient correlation presented by
McGreehan & Schotsch (1988) connected with “empirical curve fits” with compressible test
data “inter alia from Deckker & Chang (1965)”. The Hüning correlation from the same article
uses a similar approach to the Parker and Kercher correlation with the difference being the
treatment of the inlet radiusing and addition of cross-flow effects. The Idris and Pullen
correlation is also based on the correlation from McGreehan & Schotsch (1988) and the
compressibility correction comes as an extra term.
For L/D = 0.4 and r/D = 0.04, the high discharge coefficient from the current study is not
predicted by any of the correlations as observed in Figure 6.15 (a). The Parker and Kercher
correlation and the Hüning correlation best follow the trend by with a mean relative difference
69
of 8.1% and 5.5% respectively. The Idris and Pullen correlation has the same agreement in
trend even though the slope of the discharge coefficient is not as similar and has a mean
relative difference of 7.6%. For L/D = 2.0 shows that the Parker and Kercher, Hüning and
Idris and Pullen correlations follow the trend and have a mean relative difference of 3.3%,
3,1% and 3.4% respectively. The Parker and Kercher correlation and subsequently the Hüning
correlation are not valid for L/D = 4.0 and thus the comparison for L/D = 4.0 and r/D = 0.04
is made for the Idris and Pullen correlation seen in Figure 6.16. The Idris and Pullen
correlation has a mean relative difference of 3.4%.
(a) (b)
Figure 6.15: Comparison of data with correlations. (a) L/D = 0.4, r/D = 0.04. (b) L/D = 2.0, r/D = 0.04.
70
Figure 6.16: Comparison of data with correlations, L/D = 4.0, r/D = 0.04.
A summary of the characteristics of the agreement of the correlations with all the data is
presented below:
L/D = 0.4: Good agreement in trend between correlations for sharp edged orifices. The
correlations yield higher discharge coefficients than the ones obtained. For radiused
orifices, all correlations yield lower discharge coefficients compared with the
calculated ones, however the trend is in agreement with the current study.
L/D = 2.0: Good trend agreement is seen between all correlations and CFD results for
sharp-edged orifices. For radiused orifices, all correlations present a lower discharge
coefficient for low r/D and yield a higher discharge coefficient for high r/D.
L/D = 4.0: The remaining correlation has the same behaviour as seen for L/D = 2.0
The mean and maximum relative difference for all data points are presented in Table 6.4. It
can be concluded that amongst the correlation, the Hüning correlation is most in agreement
with this study, however the Idris and Pullen correlation which is also valid for L/D = 4.0 also
yields good results. The Parker and Kercher correlation in which the Hüning was improved
upon performed slightly worse than its successor.
Table 6.4: Maximum and mean relative difference of correlations.
Correlation Mean relative difference
[%]
Max relative
difference [%]
Parker and Kercher (1991) 4.7 14.3
Hüning (2008) 4.0 11.8
Idris and Pullen (2005) 3.9 15.4
71
7 Results – Study II, cross-flow
In this chapter, the results from the cross-flow study, Study II, are presented and the effect of
each parameter is presented. Moreover, comparisons will be made with previous studies with
inlet cross-flow, axially rotating orifices and with existing correlations. A summary of the
parameters is presented in Table 7.1.
Table 7.1: Summary of flow cases in Study II.
L/D [-] r/D [-] U/cax [-] p1/p2 [-]
2.0 0-0.30 0.17-1.12 1.1
4.0 0-0.30 0.17-1.12 1.1
The cross-flow ratio can be seen to range from moderate values up to very strong cross-flow
where the cross-flow velocity exceeds the ideal orifice velocity. The software in use for
secondary air system design employs static values for calculation so the cross-flow velocity
ratio is based on the ideal orifice velocity based on the static pressure ratio and the
analysis will be made with a bias towards the static discharge coefficient . The static
pressure ratio was set to 1.1 which is a common operating range for the secondary air systems
in gas turbines.
72
7.1 General overview
All the results are shown for the static discharge coefficient and the total discharge
coefficient in Figure 7.1 and Figure 7.2 respectively. It should be reminded that they both
stem from the same calculations but are just two different ways of calculating the discharge
coefficient.
It can be observed that increasing the cross-flow ratio decreases the total discharge coefficient
CD,t for all cases, but the static discharge coefficient CD,s for the highest radius to diameter
ratios r/D increases and can exceed unity. For a length to diameter ratio L/D = 2.0, the static
discharge coefficient decreases after the increase seen by the radius to diameter ratio r/D = 0.3
but its counterpart for L/D = 4.0 increases with increasing cross-flow ratio with no decrease in
the range of this study. For the same L/D, the static discharge coefficient for r/D = 0.20
increases and then decreases with increasing cross-flow ratio. It is also observed that
increasing the radius to diameter ratio r/D increases both the static and total discharge
coefficients. Elongating the orifice from L/D = 2.0 to L/D = 4.0 can be seen to be beneficial to
the static and total discharge coefficient. In the next subchapter, a more detailed analysis will
be carried out.
(a) (b)
Figure 7.1: Static discharge coefficient CD,s. (a) L/D = 2.0. (b) L/D = 4.0
73
(a) (b)
Figure 7.2: Total discharge coefficient CD,t. (a) L/D = 2.0. (b) L/D = 4.0.
7.2 Detailed analysis
This subchapter seeks out to provide a thorough explanation of the effects of the variables and
their combined effects on the static discharge coefficient CD,s.
7.2.1 L/D and U/cax
A representative radius to diameter ratio r/D = 0.08 is chosen for the analysis. It can be
observed from Figure 7.3 (a) that the static discharge coefficient is reduced with increasing
cross-flow ratio. L/D = 4.0 yields a higher static discharge coefficient than L/D = 2.0. Figure
7.3 (b) shows that for low cross-flow ratios, there seem to be no difference in static discharge
coefficient between L/D = 2.0 and L/D = 4.0 but for high cross-flow ratios, there is certainly a
benefit with increasing L/D. Figure 7.4 illustrates the separation region for L/D = 2.0, r/D =
0.08, U/cax = 0.689 and L/D = 4.0, r/D = 0.08, U/cax = 0.689. The separation region is
certainly larger than for cases without cross-flow and furthermore asymmetric. The larger
separation region reduces the through flow (Hüning, 2008) but it is observed that the flow
reattaches for L/D = 4.0. The reattachment is most likely the cause of the improvement of the
discharge coefficient.
74
(a) (b)
Figure 7.3 (a) Static discharge coefficient CD,s versus the cross-flow ratio U/cax, r/D = 0.08. (b) Static discharge coefficient CD,s versus the length to diameter ratio L/D, r/D = 0.08.
(a) (b)
Figure 7.4: Recirculation region illustrated by contour plots of negative axial velocities. (a) L/D = 2.0, r/D = 0.08, U/cax =0.689. (b) L/D = 4.0, r/D = 0.08, U/cax =0.689.
7.2.2 L/D and r/D
A representative cross flow ratio = 0.259 is chosen. Figure 7.5 (a) reveals that similarly
to the case with no cross flow, the benefits of radiusing is not as pronounced for large
radiusing. Furthermore, Figure 7.5 (b) reveals that there is a small benefit for orifices with
small or no radiusing to increase the length to diameter ratio L/D.
75
(a) (b)
Figure 7.5: (a) Static discharge coefficient CD,s versus the radius to diameter ratio r/D, U/cax = 0.259. Right: Static discharge coefficient CD,s versus the length to diameter ratio L/D, U/cax =
0.259.
7.2.3 r/D and U/cax
A length to diameter L/D = 4.0 is chosen for this analysis. Firstly, Figure 7.6 (a) is identical to
Figure 7.1 (b). The increase in the static discharge coefficient CD,s with increasing cross-flow
ratio U/cax for the highest radius to diameter r/D = 0.30 is quite remarkable. It should also be
noted that the static discharge coefficient exceeds unity, i.e. the mass flow exceeds the ideal
mass flow. The mechanism behind this is the fact that there exists a stagnation point at the
orifice inlet as seen in Figure 7.7 (a). Due to the stagnation region, there is a rise in static
pressure as evidenced in Figure 7.7 (b). The static pressure is now higher (shown as gauge
pressure in the figure) and will yield a higher pressure ratio than the original static pressure
ratio. The flow is thus driven by a pressure ratio higher than the original static pressure
equivalent to 1.1. Figure 7.8 presents the mass-flow averaged total pressure at the orifice axial
distance x/L = 0.75 and at the exit plane x/L = 1 divided with the ambient static pressure. It
can be seen that the pressure ratio is indeed larger than 1.1, i.e. some of the dynamic pressure
from the cross-flow has been redirected into the orifice. This is in line with the discussion in
subchapter 3.2.3.3 (Hüning, 2008). Figure 7.6 (b) illustrates that the benefits for increasing
radius to diameter ratio r/D are most pronounced for strong cross flow ratios.
76
(a) (b)
Figure 7.6: (a) Static discharge coefficient CD,s versus the cross flow ratio U/cax, L/D = 4.0. (b) Static discharge coefficient CD,s versus the radius to diameter ratio r/D, L/D = 4.0.
(a) (b)
Figure 7.7: Stagnation region, L/D = 4.0, r/D = 0.30, U/cax = 1.12. (a) Velocity contour plot. (b) Static pressure contour plot.
77
Figure 7.8: Pressure ratio pt,orifice/p2 at x/L = 0.75 and x/L = 1.0 for L/D = 4.0 , r/D = 0.30.
7.2.4 L/D, r/D and U/cax
Some features only manifest themselves when all three parameters are varied. Figure 7.1
shows that the static discharge coefficient increase seen for L/D = 2.0, r/D = 0.30 is followed
by a decrease where as that is not the case for L/D = 4.0, r/D = 0.30. It is believed that
increasing the cross-flow ratio sufficiently high, the static discharge coefficient will start to
decrease. It is also discovered that for strong cross-flow ratios of U/cax = 1.12, increasing the
length to diameter ratio L/D is beneficial for all r/D as illustrated by Figure 7.9. This was not
the case for weaker cross-flow ratios as seen in Figure 7.5 (b).
Figure 7.9: Static discharge coefficient CD,s versus the length to diameter ratio L/D, U/cax = 1.12
78
7.3 Comparison with data from inlet cross-flow and rotating orifices
There is a lack of data in open literature regarding cross-flow with radiused orifices. Rohde, et
al. (1969) presented two orifices with length to diameter ratio L/D = 1.06, radius to diameter
ratio r/D = 0.1951 and r/D = 0.4878 respectively and it is compared with L/D = 2.0, r/D =
0.20 and r/D = 0.30 in Figure 7.10 (a). Additionally, a comparison can be made with L/D =
4.0, r/D = 0 and is presented in Figure 7.10 (b). There are quite many differences in how
Rohde, et al. (1969) present the data. To start with, they utilize a velocity head ratio
defined as:
( 7.1 )
A transformation of the velocity head ratio to the cross-flow velocity ratio utilized in this
work can be found in the article by Hüning (2008) and is based on the ideal velocity
based on the total upstream pressure. Furthermore the data is presented for a constant main
duct Mach number and not constant static pressure ratio which is the case for the current
work. Hüning (2008) also provides a transformation law for the static pressure ratio across the
orifice and it is calculated that the static pressure ratio varies between 1.01 up to 1.6 whereas
the current study only has a static pressure ratio of 1.1. Nonetheless, the trend of a decreasing
discharge coefficient with increasing cross-flow ratio is consistent between Rohde, et al.
(1969) and the current study even no direct comparison can be made in Figure 7.10 (a) due to
the length to diameter ratio disparity. Conversely, in Figure 7.10, it can be observed that the
CFD data underpredict the discharge coefficient, a trend which has been observed in the
validation study with data from Hüning (2012).
(a) (b)
Figure 7.10: Comparison with Rohde, et al. (1969). (a) L/D = 1.06-2.0, r/D = 0.1951-0.4878. (b) L/D = 4.0, r/D = 0.
79
Jakoby, et al. (1997) studied axially rotating orifices and utilized both the absolute and
relative definition of the discharge coefficient, and respectively. The
configuration was similar to the current study with a pressure ratio p1/p2 = 1.1, L/D = 2.66 and
r/D = 0-0.5. A comparison can be made but one has to keep in mind that axially rotating
orifices and inlet cross-flow are not identical as discussed in subchapter 3.2.3.4. Figure 7.11
presents all four definitions of the discharge coefficients where the absolute discharge
coefficient CD,abs can be compared with the static discharge coefficient CD,s and the relative
discharge coefficient CD,rel can be compared with the total discharge coefficient CD,t. In Figure
7.11 (a), the trends of the current study and the work of Jakoby, et al. (1997) agree to some
extent, with decreasing discharge coefficients with increasing cross-flow ratio. However, the
current study yields consistently lower discharge coefficients and the small rise in the absolute
discharge coefficient is not present. In Figure 7.11 (b), Jakoby, et al. (1997) present an
increase followed by a decrease of the absolute discharge coefficient with increasing cross-
flow ratio. This is also seen in the current study where the static discharge coefficient
increases, exceeds unity and then decreases with increasing cross-flow ratio. However, the
maximum is shifted to higher cross-flow ratios for this case. They also experience a rising
absolute discharge coefficient for L/D = 2.66, r/D = 0.5 not followed by a decrease with
increasing cross-flow ratio. This is similar to the findings of this study for L/D = 4.0, r/D =
0.3 as seen in the Figure 7.6 (a).
(a) (b)
Figure 7.11: Comparison with Jakoby, et al. (1997). (a) L/D 2.0-2.66, r/D = 0.08-0.1. (b): L/D = 2.0-2.66, r/D = 0.30.
Dittman, et al. (2003) studied experimentally axially rotating short orifices with the largest
length to diameter ratio of L/D = 1.25. The relative discharge coefficient was found to be
independent of pressure ratio for the largest L/D = 1.25 and largest r/D = 0.5 and the data
presented here does not have a specific pressure ratio p1/p2 but is rather an average of all
pressure ratios as understood by Hüning (2008). Dittman, et al. (2003) use the cross-flow ratio
defined by the ideal velocity based on the relative total upstream values as presented
in Equation ( 3.15 ), comparable with the ideal velocity seen in the current study.
Furthermore, the relative discharge coefficient is presented which should be compared
80
with the total discharge coefficient in the current study. As in the previous case, no direct
comparison can be made between the current study and the work by Dittman, et al. (2003) due
to the additional effect experienced by rotating orifices and also due to the shorter length of
the orifice. Nonetheless, it is observed in Figure 7.12 that trends agree quite well with the
discharge coefficient decreasing with increasing cross-flow ratio. The current study does not
experience such a sharp decrease for low cross-flow ratios as present in the rotating case.
Furthermore, the discharge coefficient for the current study can be observed to be elevated
compared with the rotating orifices in most cases.
Figure 7.12: Comparison with Dittman et al. (2003), L/D = 1.25-2.0, r/D = 0-0.20.
7.4 Comparison with correlations in literature
There is only one correlation from the previous chapter that include the effect of inlet cross-
flow; the Idris and Pullen correlation (the Hüning correlation has a highest L/D = 1.25 for
inlet cross-flow and is not applicable for the current cases).
Idris and Pullen (2005) included the effect of rotation in their correlation based on
experimental studies. While an explicit upper limit of the cross-flow ratio is not stated in their
correlation, the highest cross-flow ratio presented from the experimental work is 0.53.
An extrapolation is suggested for larger values which is not included in this comparison. This
makes this correlation quite limited for comparison since the cross-flow ratio for the current
study extends to 1.12. However, Figure 7.13 shows that the trend from the correlation match
quite well with the current data even though the correlation is based on data from rotating
orifices.
81
(a) (b)
Figure 7.13: Comparison with Idris and Pullen correlation. (a) L/D = 2.0, r/D = 0-0.30. (b) L/D = 4.0, r/D = 0-0.30.
The maximum and mean relative difference can be seen in Table 7.2. From the current
analysis, it is concluded that the Idris and Pullen correlation works well but is quite limited in
range.
Table 7.2: Max and mean relative difference of correlations
Correlation Mean relative difference
[%]
Max relative difference
[%]
Idris and Pullen 3.2 6.6
82
8 Correlation
The ultimate goal with this thesis is to implement the data into a secondary air system
program which solves large complex flow networks. The data obtained in this study can be
used in its current form with interpolation for variable values in between the ones studied. It is
however more convenient to implement a correlation. Three different correlations were
created, the first one belonging to the case without cross-flow, Study I. The second and third
correlations belong to Study II and concern the static and total discharge coefficient. The
current secondary air system software uses only one of the two definitions of the discharge
coefficient but it is useful to present correlations for both for future purposes.
8.1 Correlation- Study I
There are multiple ways to correlate data, and this work utilizes a relatively simple approach.
The ultimate goal is to correlate the data to provide a discharge coefficient as a function
dependent on the three studied variables, i.e.
. The correlation is based on all the
data obtained in chapter 6, including the flow cases with a small supersonic region and the
range is presented in Table 6.1. The procedure to create the correlation can be described in
three steps:
1. Create a second degree polynomial fit for each geometry as a function of the pressure
ratio p1/p2. The result is 17 different correlations
with three coefficients for
each correlation.
2. Group the correlation into three groups dependent on the length to diameter ratio L/D.
Correlate the constants in the previously obtained correlations with each other with a
fourth degree polynomial as a function of the radius to diameter ratio r/D. Three
correlations
, one for each L/D are obtained. Each correlation has 15
coefficients.
3. Correlate the coefficients in the three correlations with each other with a second
degree polynomial as a function of the length to diameter ratio L/D and obtain a final
correlation for the discharge coefficient
with 45 coefficients.
The correlation can be written in a compact form, with all coefficients stored in the third order
tensor , as:
(
) ∑∑ ∑ (
)
(
)
(
)
( 8.1 )
The correlation is, in spite of its complex form, simple to implement in a software. The
coefficients can be found in Table C. 1. The performance of the correlation is excellent with a
maximum error compared with the data it is based on of 1.87% and a mean error of 0.3%. In
Figure 8.1, the discharge coefficients given by the correlation are plotted against the discharge
coefficients the correlation is based on. The points should ideally be located on the line but it
83
can be seen that for high discharge coefficients, the correlation tend to yield slightly higher
values compared to the discharge coefficients on which it is based on.
Figure 8.1: Discharge coefficients given by the correlation versus the discharge coefficients from the current study.
The correlation also yields good values for intermediate values of L/D, r/D and p1/p2 in almost
all cases. An example can be seen in Figure 8.2 (a) where the discharge coefficient line for
r/D = 0.02 provided by the correlation is located between the discharge coefficients lines for
r/D = 0 and r/D = 0.04. The discharge coefficient line of r/D = 0.06, which is given by the
correlation, is found between the lines from r/D = 0.04 and r/D = 0.08. There is however one
region where the correlation perform poorly. For values of L/D = 2.0-4.0 and r/D = 0.30, the
correlation is good for L/D = 2.0 and L/D = 4.0 but for L/D values in between where there are
no data points, the correlation present an unphysical behavior. Figure 8.2 (b) reveals that for
L/D = 2.7 and L/D = 3.4, the discharge coefficient is even lower than the discharge coefficient
for L/D = 4.0 and the line has a curved shape which is not present in the CFD data. As a
matter of fact, this problem extends down to r/D = 0.25. The suggestion is to limit the general
correlation to a highest radius to diameter ratio r/D = 0.2. This is then in agreement with the
shortest orifice with a length to diameter ratio L/D = 0.4 which did not include r/D = 0.30. If a
radius to diameter ratios 0.20 < r/D ≤ 0.30 are to be investigated, it is proposed that the
discharge coefficient for L/D = 2.0 and L/D = 4.0 are calculated for the desired radius to
diameter ratio r/D. Subsequently a linear interpolation can be utilized to find the discharge
coefficient for the desired L/D.
84
(a) (b)
Figure 8.2: Comparison of intermediate variable values with CFD data. (a): L/D = 2.0, r/D = 0-0.08. (b): L/D = 2.0-4.0, r/D = 0.30.
8.2 Correlations-Study II
Two sets of data were produced from the cross-flow study, one for the static discharge
coefficient and one for the total discharge coefficient . The cross-flow ratio is based
on the static pressure ratio and is thus . The creations of the correlations are based on
the same procedure presented in subchapter 8.1 and based on the data from chapter 7, in total
34 data points. The steps are identical for the total and static discharge coefficients and are
outlined for a general discharge coefficient below:
1. Create a correlation for each geometry with a third degree polynomial as a function of
. This yields 8 correlations for the discharge coefficient (
) with four
coefficients each.
2. Sort the correlations into two different groups as a function of L/D and correlate the
coefficients of the correlations in respective group with a third order polynomial as a
function of r/D. Two discharge coefficient correlations (
) with 16
coefficients are created from this step.
3. Correlate the coefficients in the two correlations with each other with a first degree
polynomial as a function of L/D, this yields a final correlation for the discharge
coefficient,
with 32 coefficients.
The correlations for the static discharge coefficient and the total discharge
coefficient can be written in a compact form as presented in Equation ( 8.2 ) and
Equation ( 8.3 ) with the coefficients stored in the third order tensor and
respectively. The coefficients can be found in Table C. 2 and Table C. 3
85
(
) ∑ ∑ ∑ (
)
(
)
(
)
( 8.2 )
(
) ∑∑ ∑ (
)
(
)
(
)
( 8.3 )
The performance of the two correlations are excellent with Equation ( 8.2 ) having a
maximum error of 3.27% and a mean error of 0.45% and Equation ( 8.3 ) having a maximum
error of 1.07 % and mean error of 0.13%. The static discharge coefficients from the
correlation are plotted against the static discharge coefficients they are based on in Figure 8.3.
The correlation is optimistic and yield slightly higher values for the values not on the line.
Figure 8.4 shows that the total discharge coefficients given by the correlation agree quite well
with the data since most of the points are located on the line. Furthermore, the correlations
work well for all intermediate variable values, e.g. Figure 8.5 shows that for intermediate
values of r/D, the correlation presents reasonable values.
Figure 8.3: Static discharge coefficients CD,s given by the correlation versus the static discharge coefficients CD,s from the current study.
86
Figure 8.4: Total discharge coefficients CD,t given by the correlation versus the total discharge coefficients CD,t from the current study.
(a) (b)
Figure 8.5: Comparison of intermediate variable values with CFD data. L/D = 2.0, r/D = 0.08-0.30 (a): Static discharge coefficient CD,s. (b): Total discharge coefficient CD,t.
8.3 Concluding remarks
An overview of the correlations and its properties can be found in Table 8.1.
87
Table 8.1: Overview of correlations.
Correlated
variable L/D r/D p1/p2 U/cax
CD 0.4-4.0 0-0.30 1.05-1.4 0
CD,s 2.0-4.0 0-0.30 1.1 0.17-1.12
CD,t 2.0-4.0 0-0.30 1.1 0.17-1.12
The first correlation for CD without cross-flow has to be used with care for 0.20 < r/D ≤ 0.30
due to several reasons. Firstly, it is not defined for L/D < 2.0 since no calculations were made
for L/D = 0.4. Secondly, it provides unphysical results in the range 0.20 < r/D ≤ 0.30 if L/D ≠
2.0 or 4.0. A possible remedy has been discussed in subchapter 8.1. The correlations for
cross-flow can be used everywhere inside the range of the variables.
All the correlations are based on polynomial functions which are meaningful inside the range
of correlated data. However, if the correlations are used outside of the specified range, the
results will not be physically meaningful or correct due to the polynomial functions.
Therefore, great care must be taken to not to use the correlations outside their range or employ
a method to deal with out of bound variables. The simplest way is to use the highest value
allowed for the variable out of bounds. More sophisticated methods include the use of
damping functions based on observations by other authors to achieve a physical meaningful
behaviour outside the range. None of these methods have been implemented in the current
work and are presented as possible future improvements. Some small suggestions based on
the observations of the current study are however given here:
If the radius to diameter ratio r/D exceeds 0.20, the discharge coefficient will
decrease for L/D = 0.4. For L/D ≥ 2.0, the discharge coefficient can be taken as
constant for r/D > 0.2.
For L/D > 4.0, the discharge coefficient can be taken to linearly decrease as seen
for 2.0 ≤ L/D ≤ 4.0.
For p1/p2 > 1.4, the discharge coefficient is dependent on the choking behavior
which is dependent on both L/D and r/D. Binder (2013) provides an excellent
overview of choking behavior for different L/D and w/D which should be similar
to the current study.
Correlations are very convenient and present an elegant way to join data points, however the
usefulness of the correlations can be questioned in this work. For the correlation for the
discharge coefficient without cross-flow , 85 data points are used and the final correlation
has 45 coefficients. It might perhaps be simpler to include all the data points and use
interpolation to find the desired discharge coefficients. Furthermore, this would avoid the
problems experienced with the correlation in chapter 8.1. This is more evident when regarding
the correlations for the discharge coefficients related to the cross-flow study. The correlations
have 32 coefficients whereas the data points only amount to 34. Thus very little has been
gained by implementing a correlation in terms of stored values.
88
9 Conclusion
This work is a continuation of the experimental work by Binder (2013) conducted at Siemens
Industrial Turbomachinery, Finspång Sweden. The objective was to, through CFD, investigate
additional parameters, both geometrical and physical, regarding flow through orifices
commonly used in secondary air systems in gas turbines.
The first goal was to investigate the effect of radiused orifices in combination with different
orifice length to diameter and pressure ratios. A verification of the computational model was
performed to ensure that the mesh was sufficient for the needs of the study and also a study of
turbulence models was conducted and compared against experimental data. To validate the
model, comparisons were made with experimental data by Binder (2013) with good
agreement, although the observed “bubble” effect could not be replicated. The flow behaviour
of sharp and radiused orifices were examined and deeper insight of the flow behaviour was
gained. The results were compared with data from chamfered orifices by Binder (2013) and
existing correlations and in general, good agreement was found. A correlation based on
polynomials was created and was found to agree well with the data it is based on although
there exists some limitations.
The second goal was to investigate the behaviour of orifices with an inlet cross-flow. No
mesh verification was made due to time constraints but a validation study was made based on
data from Hüning (2012). An analysis of the flow behaviour was conducted with the most
interesting find being the fact that dynamic pressure from the cross-flow is recovered in the
orifice giving a rise of the static pressure and in some cases an increase of the orifice through
flow. Due to the lack of data for inlet cross-flow on radiused orifices, comparisons were made
mainly with studies and correlations concerning axially rotating orifices and similar trends
could be observed. Two correlations were created for the two definitions of the discharge
coefficient presented in the current study and the performance of the correlations is
remarkably good in the range.
The study for inlet cross-flow is quite limited due to the time frame of this work. A natural
continuation point would be to study orifices with L/D = 0.4 and r/D = 0-0.30 since the
meshes already exist and requires quite little work to setup. Moreover, intermediate cross-
flow ratios should be run for existing geometries since the current study only had four to five
cross-flow ratios per geometry. Furthermore, different pressure ratios p1/p2 could also be
studied. A mesh verification study should also be performed for representative cases to ensure
the mesh is sufficient for the task at hand, something that unfortunately was not performed in
the current work.
Finally, CFD simulations should be validated by experiments if possible and it is
recommended to do so especially for short orifices (L/D = 0.4) since the values obtained are
relatively high compared to existing correlations.
89
Bibliography
ANSYS Inc, 2012. CFX-Solver Modeling Guide. In ANSYS 14.5 Help.
ANSYS Inc, 2012. CFX-Solver Theory Guide. In ANSYS 14.5 Help.
ANSYS Inc, 2012. FLUENT Theory Guide 14.5. In ANSYS 14.5 Help.
Binder, C., 2013. Experimental Diploma Thesis. Diploma Thesis. Finspång/Karlsruhe:
Siemens Industrial Turbomachinery/Karlsruher Institut für Technologie.
Binder, C. et al., 2014. Experimental Study on Pressure Losses in Circular Orifices for the
Application in Internal Cooling Systems, GT2014-25619. In ASME Turbo Expo 2014: Turbine
Technical Conference and Exposition. Düsseldorf, 2014. ASME.
Brain, T.J.S. & Reid, J., 1975. Performance of Small Diameter Cylindrical Critical-Flow
Nozzles. National Engineering Laboratory Report 546.
Çengel, Y.A. & Cimbala, J.M., 2006. Fluid Mechanics, Fundamentals and Applications. New
York: McGraw-Hill.
Deckker, B.E.L. & Chang, Y.F., 1965. An Investigation of Steady Compressible Flow
Through Thick Orifices. Proc Instn Mech Engrs 1965-66, 180, pp.312-23.
Dittman, M., Dullenkopf, K. & Wittig, S., 2003. Discharge Coefficients of Rotating Short
Orifices With Radiused and Chamfered Inlets, GT2003-38314. In ASME Turbo Expo 2003
Power for Land, Sea and Air. Atlanta, 2003. ASME.
Hay, N. & Lampard, D., 1996. Discharge Coefficient of Turbine Cooling Holes: A Review.
ASME Paper 96-GT-492. In Internation Gas Turbine and Aeroengine Congress & Exhibition.
Birmingham, 1996. ASME.
Hay, N. & Spencer, A., 1992. Discharge Coefficients of Cooling Holes With Radiused and
Chamfered Inlets. Journal of Turbomachinery, 114, pp.701-06.
Hüning, M., 2008. Comparison of Discharge Coefficient Measurements and Correlations for
Several Orifice Designs With Cross-Flow and Rotation Around Several Axes, GT2008-
50976. In ASME Turbo Expo 2008: Power for Land, Sea and Air. Berlin, 2008. ASME.
Hüning, M., 2012. Discharge Coefficient Measurements of Round, Inclined Orifices with Inlet
Cross-Flow in and Against Direction of Inclination. Journal of Mehcanical Engineering
Science, 227(6), pp.1266-1279.
Idris, A. & Pullen, K.R., 2005. Correlations for the Discharge Coefficient of Rotating Orifices
Based on the Incidence Angle. Proc. IMechE Vol. 219 part A: J. Power and Energy, pp.333-
52.
Jakoby, R., Geis, T..K.S. & Wittig, S., 1997. Discharge Coefficients of Rotating Orifices with
Radiused Inlet Corners. In Thirteenth International Symposium on Airbreathing Engines
(ISABE XIII). Chattanooga, Tennessee, 1997. American Institute of Aeronautics and
Astronautics.
90
Johansson, A.V. & Wallin, S., 2012. Turbulence Lecture Notes. Stockholm: KTH.
Kundu, P.K. & Cohen, I.M., 2008. Fluid Mechanics. 4th ed. Oxford: Elsevier.
Lechner, R., Langtry, R. & Menter, F., 2010. SST Reattachment Modification in ANSYS CFX.
Presentation. Otterfing: ANSYS Germany.
Lichtarowicz, A., Duggins, R.K. & Markland, E., 1965. Discharge Coefficients for
Incompressible Non-Cavitating Flow through Long Orifices. Journal of Mechanical
Engineering Science Vol 7 No 2, 7(2), pp.210-19.
McGreehan, W.F. & Schotsch, M.J., 1988. Flow Characteristics of Long Orifices with
Rotation and Corner Radiusing. Journal of Turbomachinery, 110, pp.213-17.
Parker, D.M. & Kercher, D.M., 1991. An Enhanced Method to Compute the Compressible
Discharge Coefficient of Thin and Long Orifices with Inlet Corder Radiusing. Heat Transfer in
Gas Turbine Engines, 188, pp.53-63.
Pope, S.B., 2000. Turbulent Flows. 1st ed. Cambridge: Cambridge University Press.
Rohde, J.E., Richards, H.T. & Metger, G.W., 1969. Discharge Coefficients for Thick Plate
Orifices with Approach Flow Perpendicular and Inclined to the Orifice Axis. Technical Note D-
5467. Cleveland: National Aeruonautics and Space Administration Lewis Research Center.
Schlichting, H. & Gersten, K., 2006. Grenzschichttheorie. 10th ed. Berlin, Heidelberg:
Springer-Verlag.
Tannehill, J.C., Anderson, D.A. & Pletcher, R.H., 1997. Computational Fluid Mechanics and
Heat Transfer. 2nd ed. Washington, DC: Taylor & Francis.
Ward-Smith, A.J., 1979. Critical Flowmetering: The Characteristics of Cylindrical Nozzles
with Sharp Upstream Edges. Int. J. Heat & Fluid Flow, 1(3), pp.123-32.
Weir, A.J., York, J.L. & Morrison, R.B., 1956. Two-and Three-Dimensional Flow of Air
Through Square-Edged Sonic Orifices, ASME Paper No. 54-A-112. Transactions of the
ASME, pp.481-88.
Versteeg, H.K. & Malalasekera, W., 2007. An Introduction to Computational Fluid Dynamics.
2nd ed. Essex: Pearson Education Limited.
91
Appendix A
Table A. 1: Overview of the meshes used in chapters 6 and 7.
L/D r/D Hemisphere
I elements
Hemisphere
II elements
Inner
geometry
elements [-]
Cross-flow
cube
elements [-]
0.4 0 66305 175986 1210464 -
0.4 0.04 - 163976 910848 -
0.4 0.08 - 163976 1069824 -
0.4 0.16 - 163976 1140480 -
0.4 0.20 - 163976 1193472 -
2.0 0 105872 281879 1315008
2.0 0 - 953756 4914784 -
2.0 0.04 - 163976 1259712 -
2.0 0.08 - 163976 1325952 -
2.0 0.08 - 953756 1325952 190894
2.0 0.16 - 163976 1387776 -
2.0 0.20 - 163976 1414272 -
2.0 0.20 - 953756 1414272 190894
2.0 0.30 - 163976 1555584 -
2.0 0.30 - 953756 1555584 190894
4.0 0 105872 281879 1685664 -
4.0 0 - 953756 6011464 -
4.0 0.04 - 163976 1250880 -
4.0 0.08 - 163976 1325952 -
4.0 0.08 - 953756 1325952 190894
4.0 0.16 - 163976 1467264 -
4.0 0.20 - 163976 1612992 -
4.0 0.20 - 953756 1612992 190894
4.0 0.30 - 163976 1749888 -
4.0 0.30 - 953756 1749888 190894
92
Appendix B
Figure B. 1: Axial velocity profile across the orifice at x/L = 0.476. L/D = 1.5, r/D = 0.
Figure B. 2: Turbulent kinetic energy profile across the orifice at x/L = 0.476. L/D = 1.5, r/D = 0.
93
Figure B. 3: Axial velocity profile across the orifice at x/L = 0.6. L/D = 0.4, r/D = 0.04.
Figure B. 4: Turbulent kinetic energy profile across the orifice at x/L = 0.6. L/D = 0.4, r/D = 0.04.
94
Figure B. 5: Axial velocity profile across the orifice at x/L = 0.8. L/D = 0.4, r/D = 0.04.
Figure B. 6: Turbulent kinetic energy profile across the orifice at x/L = 0.8. L/D = 0.4, r/D = 0.04.
95
Appendix C
Table C. 1: Coefficients aijk.
i = 0 k = 0 k = 1 k = 2
j = 0 0.377 0.3023 -0.05472
j = 1 1.848 -2.295 0.3039
j = 2 -126.6 89.97 -14.02
j = 3 -660.2 568.9 -99.44
j = 4 5391 -4204 707.8
i = 1 k = 0 k = 1 k = 2
j = 0 0.2052 -0.1729 0.03631
j = 1 -0.3961 5.785 -0.9304
j = 2 276.4 -228.5 37.28
j = 3 419.8 -339.2 61.26
j = 4 -7364 5670 -953
i = 2 k = 0 k = 1 k = 2
j = 0 -0.003021 0.02311 -0.006889
j = 1 0.9399 -2.989 0.4935
j = 2 -126.6 103.9 -17.06
j = 3 -75.91 66.37 -12.81
j = 4 2815 -2171 365
Table C. 2: Coefficients bijk.
i = 0 k = 0 k = 1
j = 0 0.7406 0.01075
j = 1 3.41 -0.2585
j = 2 -16.79 1.085
j = 3 27.64 -1.825
i = 1 k = 0 k = 1
j = 0 -0.1719 0.0151
j = 1 4.802 -1.302
j = 2 -25.1 8.747
j = 3 26.33 -12.42
i = 2 k = 0 k = 1
j = 0 -0.08821 0.02515
j = 1 -19.41 4.762
j = 2 157.7 -36.5
j = 3 -254.8 58.66
i = 3 k = 0 k = 1
j = 0 -0.06188 0.007975
j = 1 11.75 -2.985
j = 2 -114.5 26.74
j = 3 206.2 -46.58
96
Table C. 3: Coefficients cijk
i = 0 k = 0 k = 1
j = 0 0.742 0.01075
j = 1 3.662 -0.3215
j = 2 -19.15 1.63
j = 3 32.06 -2.815
i = 1 k = 0 k = 1
j = 0 -0.1821 0.0148
j = 1 2.583 -0.7468
j = 2 -2.904 3.699
j = 3 -15.7 -3.24
i = 2 k = 0 k = 1
j = 0 -0.4827 0.02135
j = 1 -16.2 3.62
j = 2 108.5 -24.34
j = 3 -154.6 35.59
i = 3 k = 0 k = 1
j = 0 0.2084 -0.0064
j = 1 10.25 -2.382
j = 2 -84.62 18.81
j = 3 140.3 -30.7
Recommended